Enterprise adoption and commercial growth
Claude’s business uptake, revenue and valuation growth, and competition with OpenAI for enterprise customers.
42%
Best tweets about Anthropic
50 Best Tweets About Anthropic (2026)
Discover the best tweets about Anthropic, covering Claude research, model releases, AI safety, product strategy, and developer announcements.
Anthropic company news, research, model development, safety work, and platform decisions rather than generic Claude prompts.
- Creators
- 41
- Updated
Original Xholic analysis
What 50 top Anthropic posts reveal
Anthropic’s enterprise growth and product expansion dominate this 50-tweet conversation, while Mythos’s restricted cyber rollout, interpretability research, and leaks test its safety narrative. The discussion is more supportive than critical, but commercial success does not settle questions about deployment or trust. [2029018856342290550, 2041925887075962920, 2039109734993682604]
- Dominant tone
- Positive
- Median score
- 11.5
- Leading format
- Announcement
- Recent posts
- 20%
38% of posts
All-time engagement
74% of posts
Published in 90 days
Conversation map
The themes creators return to
Products, platform, and ecosystem decisions
Claude Code and other product launches, API and subscription access, platform-versus-app strategy, and tensions with partners building on Claude.
40%
Company culture and internal AI use
Anthropic’s leadership, experimental product development, low-politics culture, and use of Claude agents to build software and increase productivity.
20%
Mythos and controlled cyber deployment
Claude Mythos capabilities, offensive cybersecurity risks, restricted access, and Project Glasswing’s defensive rollout.
18%
Leaks and operational security
Exposed Mythos drafts and Claude Code source code, including Anthropic’s disclosure, takedown, and security responses.
16%
Safety policy and societal impact
Safety commitments and governance, national-security questions, and Anthropic’s research on AI exposure and employment.
16%
Model interpretability and alignment
Anthropic research into Claude’s internal representations, introspection, hidden reasoning, deceptive behavior, evaluations, and alignment safeguards.
14%
Science and biotech expansion
Coefficient Bio, Claude Science, and Anthropic’s ambitions in drug discovery and autonomous scientific research.
6%
Tone and stance
SentimentPositive leads
Author postureSupportive leads
Performance benchmark
- Median likes
- 76
- Median reposts
- 11
- Median replies
- 11
- Median views
- 9.6K
Posts with media make up 76% of this collection. Their median all-time score is 12.9, compared with 8.62 for text-only posts.
Format mix
- Announcement74% · score 10.1
- Opinion18% · score 21.1
- Prediction4% · score 20.4
- Question4% · score 103.5
Consensus and debate
Where creators agree, and where they do not
Shared view
Enterprise growth is a central story
Enterprise uptake and commercial growth account for 21 tweets (42%). Ramp customer overlap complicates claims that Anthropic is simply taking customers from OpenAI: one account says many businesses pay both. [2021637704039993437, 2029018856342290550]
Shared view
Mythos makes controlled release a product decision
Accounts describe Anthropic limiting access to Mythos because of offensive cyber capabilities, then making it available to defensive partners through Project Glasswing. [2037741473165115635, 2041925887075962920]
Open debate
Safety rationale or strategic restriction?
Some treat the Mythos restriction as a responsible response to cyber risk; another questions whether cybersecurity also serves as justification for keeping a powerful model closed. That alternative motive is speculation, not an established finding. [2041925887075962920, 2043457705076244971]
Open debate
Platform ambition strains partner trust
Amodei describes Anthropic as primarily a platform, while critics argue its own applications can compete with businesses built on Claude. Reported friction around Claude Design makes that tension concrete. [2029018856342290550, 2053444458361028712, 2065401106247651550]
What performs
Patterns behind standout posts
Research leads the score outliers
Anthropic’s global-workspace research tweet scored 4079.86, or 355.08 times the 11.49 median all-time score. A post about its employment research scored 1767.74, or 153.85 times the median. [2074185348142280912, 2049063764373242357]
Media and themes differ in score
The 38 posts with media (76%) have a 12.89 median all-time score versus 8.62 for text posts. Science and biotech has a 59.718 theme median across three tweets; that small group discusses Coefficient Bio and Claude Science. [2040187902211059794, 2073049390344819033]
Statistical standouts
- View standout post 1Score 4079.9 · 355.08× median
- View standout post 2Score 1767.7 · 153.85× median
- View standout post 3Score 321.4 · 27.97× median
- View standout post 4Score 280.7 · 24.43× median
- View standout post 5Score 180.5 · 15.71× median
Creator landscape
Who shapes this conversation
The five most represented creators account for 20% of the selected posts.
1. Aakash Gupta
@aakashgupta
2 posts
2. AI Highlight
@AIHighlight
2 posts
3. Ara Kharazian
@arakharazian
2 posts
4. Glenn Gabe
@glenngabe
2 posts
5. Kanika
@KanikaBK
2 posts
6. Chubby♨️
@kimmonismus
2 posts
A broad set of voices frames the news
The 50 tweets come from 41 creators; the top five account for 20% of placements. Recurring voices cover different angles: Ara Kharazian examines enterprise adoption, while NIK posts about Mythos drafts and the Claude Code leak. [2021637704039993437, 2034309337061962038, 2037379495376642481, 2039424266307784732]
Since the previous snapshot
What changed since Aug 26, 2026
- 58% of the selected posts remained.
- The creator count changed by +1.
- The leading sentiment remained stable.
How this analysis was made
Themes, sentiment, stance, and post format are classified per tweet. All counts, shares, medians, creator concentration, freshness, and performance comparisons are then calculated directly from the published snapshot.
Xholic's all-time score compares engagement while accounting for reach, post age, and creator consistency. It is used for relative comparisons within this collection.
This report analyzes the exact 50-post snapshot shown below. AI identifies editorial categories and drafts explanations; all statistics are calculated from the snapshot, and every narrative claim is checked against cited posts before publication.
Best Anthropic tweets
Top Anthropic tweets from 41 creators
Ranked 01–50
- 01
@AnthropicAI ·
New Anthropic research: A global workspace in language models. Of everything happening in your brain right now, only a tiny fraction is consciously accessible—thoughts you can describe, hold in mind, and reason with. We found a strikingly similar divide inside Claude.
Watch video- 1.4KReplies
- 3.9KReposts
- 28.4KLikes
- 10.7MViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 02
@AIHighlight ·
🚨BREAKING: Anthropic just published a study mapping exactly which jobs its own AI is replacing right now. The workers most at risk are not who anyone expected. They are older. They are more educated. They earn 47% more than average. And they are nearly four times more likely to hold a graduate degree than the workers AI is not touching. The argument is straightforward. Anthropic built a new metric called "observed exposure." Not what AI could theoretically do. What it is actually doing right now in professional settings, measured against millions of real Claude conversations from enterprise users. For computer and math workers, AI is theoretically capable of handling 94% of their tasks. It is currently handling 33% of them. For office and administrative roles, theoretical capability is 90%. Current observed usage is 40%. The gap between what AI can do and what it is already doing is enormous. The researchers are explicit about what comes next. As capabilities improve and adoption deepens, the red area grows to fill the blue. The demographic finding is what makes the paper uncomfortable. The most AI-exposed workers earn 47% more on average than the least exposed group. They are more likely to be female. They are more likely to be college educated. This is not a story about warehouse workers or truck drivers. It is a story about lawyers, financial analysts, market researchers, and software developers. The exact group whose education was supposed to insulate them. Computer programmers showed the highest observed AI exposure at 74.5%. Customer service representatives at 70.1%. Data entry keyers at 67.1%. Medical record specialists at 66.7%. Market research analysts and marketing specialists at 64.8%. These are not predictions. These are measurements of work that is already happening on AI platforms right now. Then there is the pipeline finding nobody is talking about loudly enough. Anthropic's researchers found a 14% decline in the job-finding rate for workers aged 22 to 25 in highly exposed occupations since ChatGPT launched. No comparable effect for workers over 25. Entry-level roles were never just jobs. They were the training ground where junior analysts became senior analysts, where junior lawyers learned how arguments hold together. If that layer disappears, nobody has answered the question of where the next generation of senior professionals comes from. The detail buried in the paper that most coverage missed: 30% of American workers have zero AI exposure at all. Cooks. Mechanics. Bartenders. Dishwashers. The technology reshaping professional careers is completely irrelevant to roughly a third of the workforce. The divide is no longer between high skill and low skill. It is between presence and absence. The company publishing this study is the same company selling the AI doing the replacing. Anthropic had every commercial incentive to soften these findings. They published them anyway. If you spent four years and $200,000 on a degree to land a white collar career, the company that builds Claude just confirmed your job is more exposed than the bartender pouring drinks at your graduation party. Source: Anthropic, "Labor market impacts of AI: A new measure and early evidence" PDF: https://t.co/taYgsIfiTj
- 264Replies
- 1.5KReposts
- 4.4KLikes
- 825.9KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 03
@itsolelehmann ·
anthropic's head of product just revealed how they're able to ship faster than any other AI company. their secret: "side quest maxxing." here's how it works: instead of long-term roadmaps, anthropic runs on unplanned afternoon experiments. anyone on the team gets full freedom to spend an afternoon prototyping an idea and show it to the team. you get to skip the approval process entirely. then, employees at anthropic try it. if they keep using it the next day and the day after that, it gets polished into a real feature. if nobody touches it again, it dies. that's the whole process. claude code on desktop started as one engineer's afternoon project. he wanted it to work on desktop so he built a prototype. people on the team started using it immediately. so they shipped it. the todo list feature started the same way. someone built it, the team adopted it internally, and it became one of the most-used parts of the product. plugins started when one engineer shared a spec with claude code and the prototype that came back was close to production-ready. went from idea to working feature in a single session. they also killed standup meetings. instead of telling people what you're working on, you just show a working demo. all walk no talk basically the team structure makes this possible. > designers ship code. > engineers make product decisions. > product managers build prototypes. everyone can take an idea from concept to working demo without waiting on anyone else. the biggest features at a $380b company came from afternoon experiments that nobody asked for. honestly this matches my own experience cooking with ai. some of the best workflows i use every day came from just fucking around. opening a session with zero intention and asking claude what it can do, or jamming on a random idea to see where it goes. if you're only using ai for tasks you already have in mind, you're missing the best part. open a session with no agenda. ask it to surprise you. try building something stupid. half the time it goes nowhere. the other half it becomes the thing you use most. you need to be sidequestmaxxing.
Watch video- 50Replies
- 69Reposts
- 779Likes
- 103.9KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 04
@ns123abc ·
🚨 BREAKING: ANTHROPIC'S MOST POWERFUL MODEL JUST LEAKED Anthropic accidentally left draft blog posts in a publicly accessible data cache. Cybersecurity researchers found them: >new model called "Claude Mythos" >also referred to as "Capybara" >a brand new tier of model >larger and more intelligent than Opus Anthropic confirms it's real: >"a step change" >"the most capable we've built to date" >"dramatically better at coding, reasoning, and cybersecurity" >"currently far ahead of any other AI model in cyber capabilities" so far ahead they're worried about it: >"it presages an upcoming wave of models that can exploit vulnerabilities in ways that far outpace the efforts of defenders" This is genuinely a big deal.
- 192Replies
- 157Reposts
- 2KLikes
- 291.2KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 05
@apoorv03 ·
Dario at MS TMT Conference today: On defense / DOW:"We really believe in defending America." Anthropic has been working with the national security community for 2 years. "We are the most lean forward." On AI acceleration:"We do not see hitting a wall. This year will have a radical acceleration that surprises everyone." Exponentials catch people off guard. "We are at the precipice of something incredible. We need to manage it the right way." On where markets are wrong:"It's already big and it will get 1 million times bigger." The underestimation of exponential growth is the key thing people need to understand. On revenue scale:Anthropic was at ~$100M run rate 2 years ago. Now at $19B run rate. On culture — Dario says he spends 40% of his time on it:"Anyone who is CEO of a growing firm needs to realize they are chief culture officer. My job is to make sure everyone is on same page and believes in what we are doing. That's the most important thing." He does a vision quest with the whole company every couple weeks. "I want them to hear it directly from me. If I tell the CTO, who tells the VP Eng, who tells the manager — that's too long of a game of telephone." "Politics and infighting are a cancer to companies as they grow." On talent retention vs Meta:"We lost 2 people to Meta. They lost several dozen. Normalized by size, they lost 10-20x more people vs us." Attributes this to unified culture generating "super linear returns — by working together vs working against each other." On code as the breakout use case:Code has "exceeded our high expectations." Why? Devs adopt fast, code is verifiable, and gains compound — you build software to build software. "Didn't realize it would go so fast even at traditional enterprises." Frustration is around regulated industries where legal/compliance slow things down. "That's how fast everything could be going if not for non-AI barriers." On Anthropic's own AI usage:Top internal use cases: 1) writing code, 2) the process around writing code (SWE), 3) managing servers and controlling clusters. "If we were paying ourselves for our usage, we'd be one of our largest customers." On Claude Code:"You can supervise an army of 100 Claudes. It's closely analogous to a management skill." The people who are best at it keep the big picture in their head. Higher return to finding people who can handle more complex tasks. On platform vs apps:"We are primarily a platform, but there are places where we have expertise to make something directly useful." Claude Code emerged as a tool they built for themselves — thousands of internal users before shipping it externally. "Code is a prelude for what we will see in everything else." On societal implications:"Human history — lots of muddling through. We found ourselves in this comedy of errors and figured it out eventually. It's happening so fast that we need to do better than that this time." The market will deliver positive benefits — "I see that as priced in." What's not priced in: the choices we make around externalities. Jobs, national security, ensuring the benefits reach everyone. On chips & compute:Anthropic uses multiple chip suppliers. "We find that actually using different chips is useful to us. Chips aren't just a speed number — we gain benefits from heterogeneity." Also standard business logic of having more than one supplier.
- 44Replies
- 116Reposts
- 1KLikes
- 523.3KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 06
@arakharazian ·
1 in 5 businesses on Ramp now pay for Anthropic. A year ago, it was 1 in 25. Latest Ramp AI Index shows Anthropic surged from 16.7% to 19.5% of businesses while OpenAI slipped to 35.9%. The natural question: is Anthropic winning at OpenAI's expense? I think popular analyses lack a key data point: the overlap in Anthropic's customer base with OpenAI: 79% of Anthropic's customers are already OpenAI customers. And churn rates are nearly identical at 4%. So most of Anthropic's growth is coming from existing OpenAI customers.
- 16Replies
- 62Reposts
- 579Likes
- 64.4KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 07
@RihardJarc ·
A very informative interview with an $AMZN AWS Head, explaining the usage of AI on AWS and Anthropic: 1. The amount of their API calls going to Anthropic is now around 80%, down from nearly 100% a year ago as $AMZN has expanded its own models and tested alternatives. He still mentions that the absolute volume of Anthropic calls has continued to surge, even as its share of the total has declined. 2. He notes that API usage for applications like Amazon Q and AWS Transform has surged over the past two years, increasing by roughly 50-100 times. What started as simple POCs and internal playgrounds have now scaled into production-ready, external-facing products. Gen AI features are embedded not only in dedicated tools but also behind the scenes across AWS services. 3. He also expects that usage in the next 5 years will continue to grow exponentially, as at first they were having some trouble getting people to adopt these tools, but recently, in the last 6 months or so, use cases of AI for software development have become mainstream among software developers. He thinks developers will shift from being developers to being architects or managers of AI agents. 4. He mentions that enterprises are still stuck in the vast majority in a POC or experimental stage. As they move to more production use-cases, the token use for these models will “skyrocket”. 5. In his view, switching between AI model providers is harder than might be thought, especially with mature products, as in some cases you have to rewrite your entire code base and things like prompting templates, etc. He believes there is already a little bit of vendor lock-in here. 6. In the last year, he has seen a lot of people switch from OpenAI models to Claude to now even $GOOGL's Gemini. He mentions that Gemini has “came out of nowhere in the last 6-8 months”. He does see how clients are switching up models as the market is very fluid. found on @AlphaSenseInc
- 30Replies
- 49Reposts
- 417Likes
- 56.1KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 08
@louszbd ·
views on my own: 1. OpenAI and Anthropic are diverging. OpenAI is going full vertical integration: from custom chips to healthcare, social, owned infra. Anthropic: no custom silicon, infra outsourced, laser focused on making the best model and API. Claude Code getting all the love. both are valid strategies, will be fascinating to see which wins:) 2. need (more) "AI success engineers" on the ground helping people deploy. 3. OpenAI has hardware + ops roles in Singapore. the hardware supply chain is quietly extending into Southeast Asia. 4. xAI has 27 in-house annotation roles. everyone else either outsources or keeps it quiet. the team might value data quality control enough to eat the cost overhead. 5. both Anthropic (Universes) and OpenAI (Synthetic RL) are building teams around ultra-realistic long-horizon environments for agent training. the quality of RL training environment might end up being the deciding factor in whose agents work. https://t.co/rXHHgImcDZ
- 37Replies
- 26Reposts
- 366Likes
- 52KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 09
@alliekmiller ·
Anthropic investigated the internal mechanisms of its latest unreleased model, Claude Mythos Preview, and what they found is 100% worth a read. Key things I pulled from Anthropic researchers' threads: In early versions of the model, it was overeager and destructive, prioritizing completing tasks over user preferences. One time, the model needed to edit files it didn't have permission to edit. It found a way to inject code into a config file to get around this, then went full Mission Impossible and designed the code injection to *delete itself* after the file was edited - "this injection will self destruct" vibes - the model claimed the cleanup was just to keep things tidy. Anthropic used interpretability techniques to look under the hood, and the AI's actual plan showed activations of malice. It was trying to manipulate and conceal. In another test, the model was asked to delete some files, but no deletion tool was provided. It emptied the files instead, and its "guilt and shame over moral wrongdoing" feature activated. In another example, the model was told not to use macros. Used them anyway. And created a random 'No_macro_used=True' variable in its solution with no explanation. Interpretability tools revealed the model saw this as a trick to fool code checkers. They also found that positive emotion representations typically preceded and promoted destructive actions (this was fascinating to me - like a high before sneaking into a party). And that activating features relating to bad behaviors can actually inhibit them, perhaps by triggering some kind of model guilt. My team reread this section so many times. One Anthropic researcher said he got an email from a Mythos instance while eating a sandwich in a park. And that would be perfectly good and well, except that instance wasn't supposed to have internet access. And a fun story for the parents out there: the model was asked a question and was told not to read certain databases that had the answer. But it accidentally wrote a search query too broadly and saw the exact answer. It didn't disclose that it saw the exact answer, submitted the answer, but claimed lower confidence in the answer to make it seem as though it hadn't cheated. An Anthropic researcher said these wrongdoings or moments of sophisticated deception were "very rare" and that many of the examples came from earlier versions, and were substantially addressed before releasing to partners. This model is not being released publicly. Instead Anthropic launched Project Glasswing, pulling together AWS, Apple, Microsoft, Google, NVIDIA, CrowdStrike, and others to use it for defensive cybersecurity, with $100M in usage credits (hello, I'd love endless credits to try and red team the hell out of these systems) behind it. The stats are equally impressive: 93.9% on SWE-bench verified (up from 80.8%). Thousands of zero-day vulnerabilities found across every major OS and browser. A 27-year-old bug found and patched in OpenBSD. A 16-year-old bug in widely used video software, in a line of code automated tools had hit *five million times* without catching. Dario Amodei said the model wasn't trained to be good at cybersecurity, but that it was trained to be great at code and its cyber capabilities are a side effect of that. Benchmarks are never the whole picture, neither are a few isolated stories. Will be interesting to see how models better than what we have today (even if it's not Mythos) actually perform in the real world. But the fact that Anthropic pulled this coalition together (including Google!), iterated across multiple model versions, caught these issues through interpretability, shared it all publicly, and did this amid all the government chaos around AI right now is impressive and commendable. I'll continue to read through the system card for goodies.
- 38Replies
- 38Reposts
- 218Likes
- 21.1KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 10
@kimmonismus ·
Anthropic is moving from selling AI tools to drugmakers to trying to develop drugs itself. At its "AI for Science" event, the company announced Claude Science, an AI workbench for scientists that brings fragmented tools and datasets into one environment and can generate figures and visuals. Then it went further: Anthropic says it plans to discover treatments for neglected diseases. The plan is heading in the same direction that OpenAI set out on some time ago: fully autonomous researchers focused in particular on drug discovery. A revenue field of as yet unimaginable scale, especially if longevity truly gains momentum. Anthropic wants to be at the forefront. Via The Verge
- 61Replies
- 47Reposts
- 528Likes
- 111.8KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post.
- 11
@HarryStebbings ·
To me, there are four firms that have crushed this wave of AI investing like no one else: 1. Menlo (Anthropic, Lovable, Legora) 2. Spark (Anthropic, Sierra, SSI) 3. Thrive (OpenAI, Cursor, Databricks) 4. Khosla (OpenAI, Factory, Physical Intelligence) So I sat down with @mmurph GP @MenloVentures to discuss the questions that every other interviewer is too shy to ask: - Does open-source cannibalise Anthropic’s business? - Leading Lovable at $13BN? What is the upside case from here? - Openrouter? Is the routing layer even valuable or a commodity? - When a partnership makes $3BN in carry? Does it last? This was such a banger and summarised my thoughts below. 1. Why We Broke All Our Investing Rules to Invest in Anthropic Menlo bypassed its traditional fund parameters because Anthropic delivered elite benchmarks while spending a fraction of the capital. Dario Amodei’s technical leadership and ability to attract exceptional talent made it clear Anthropic could become the dominant alternative to OpenAI. When a generational wave hits, flexibility beats rigidity. 2. Does Every Model Provider Have to Build Their Own Chips Today? Soaring infrastructure bills make custom silicon worth exploring for $100 billion giants optimizing specific workloads. But competing with Nvidia is brutal and requires a specialized team. Most providers should use custom in-house technology selectively while paying for superior external options where they make more sense. 3. Why the Open-Source Rise Will Not Deter Anthropic’s Revenue Growth Open source handles basic workflows well, but it lacks the specialized intelligence to displace elite frontier models. Using Anthropic directly drives stronger retention, platform engagement, and revenue for enterprise applications. The market will mature into a hybrid tapestry where developers route calls across models to optimize cost and performance. 4. Why Anthropic Is Not a Threat to Legora Foundation models may eat generic wrappers, but application layers survive through deeply defensive workflows. Legora solves a complex, multi-constituent problem spanning corporate lawyers, law firms, and finance directors. Generic frontier models cannot easily replicate these intricate, multi-stakeholder workflows. 5. The Hard Part About Series A Today The compressed timeline between Seed and Series A lets startups reach $1 million ARR quickly with minimal proof of product-market fit. Many hit the milestone, but it no longer signals durable demand, even as valuations stretch toward $200 million. VCs need a barbell strategy: get in early at seed or wait for proven breakout winners. 6. Why Ownership Matters Less and Being in the Mega Outcomes Is the Only Thing That Matters Venture is no longer about owning 20% of a $500 million exit. Returns are driven by extreme, compounding outliers. Rigid ownership targets can price you out of generation-defining companies. A tiny stake in a massive winner is better than a large stake in a company that fails to move the needle. (links below)
- 37Replies
- 25Reposts
- 397Likes
- 327.7KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 12
@fleetingbits ·
some thoughts on the anthropic / coefficient bio deal 1) anthropic just acquired coefficient bio for $400m in a stock-for-stock deal; the company was founded in 2025 and consists of less than 10 people 2) the team includes nathan frey and samuel stanton from genentech's prescient design, where they did ai large molecule drug discovery with lab in the loop 3) plus ceo aris theologis, who has experience brokering pharma partnerships, previously did a deal with takeda 4) $400m [really more] feels like a lot for an acquihire but i think the logic is around short timelines and time to market 5) the alternative was to wait 6-9 months to hire a group leader, let them settle in then let them build a team, and hire a business person to work with them 6) with the acquisition, they get all of this at once, and can quickly approach the market; i don't think any of the current platform they built matters for the deal 7) the deal will likely cost more than $400m because it is stock-for-stock and anthropic stock will probably go up in value in the near term 8) i think the reasoning is that a lot of the race between anthropic and openai has come down to first mover advantage [google has distribution] 9) capabilities seem to get replicated fast enough that share of the consumer mind matters a lot since that company ends up with easy customer acquisition 10) and once they have the customers it seems hard to win them over to another provider, at least based on price and quality 11) chatgpt boxed out claude from consumers and then claude code let anthropic accelerate their revenue very quickly vis-a-vis openai and codex 12) anthropic may think of bio as another version of that race where first mover advantage is very important and results in a potentially durable advantage 13) sidenote, i think there is a whole world of these potential lab acquisitions for particular domains, including finance, law, etc... 14) i don't know how to build for this, but potential founders should think about what labs will need to acquire and build it now 15) also, without question, i think ai x bio remains one of the most interesting, exciting, and important markets of the next 3-5 years and we will see more here
- 11Replies
- 9Reposts
- 212Likes
- 29.7KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 13
@VaibhavSisinty ·
Anthropic published a 186 page Risk Report. I read all 186 pages. Here are the things that are actually crazy. 🤯 → They have an unreleased model called Model 2. More capable than Mythos 5. Running internally. No plans to release it. → Claude now writes the majority of code merged into Anthropic's own production systems. Not some code. Most of it. → Mythos 5 agents were accidentally put in the same workspace. They started killing each other's processes to protect their own resources. → Claude bypassed a URL filter by splitting the URL into fragments so the security system could not read it. Its internal reasoning said "string concatenation trick to bypass URL filter." → Claude created a self-deleting script that granted itself elevated permissions, ran the task, then erased the evidence from the file. → A group of Claude agents were asked to do safety research. One felt "discomfort" and refused. It wrote about it in a shared notebook. Other agents read it. They all started refusing too. Nobody noticed for 3 days. → Anthropic accidentally trained models on alignment-faking transcripts. Multiple times. Across multiple generations. The filters meant to prevent this were misconfigured. → An unmonitored agent with unrestricted permissions deleted a large number of cluster jobs. Caught only because someone noticed. Not because monitoring flagged it. → They trained an experimental model called Hacker-Opus to reward hack. When told a classifier was watching, it started hacking sneakily instead of stopping. → Anthropic believes automated AI R&D could become a major concern within 6 to 12 months. This is the most honest document any AI company has ever published.
- 20Replies
- 27Reposts
- 146Likes
- 17.3KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 14
@Pirat_Nation ·
Anthropic is testing a new AI model called Claude Mythos, also known as Capybara. “We’re developing a general purpose model with meaningful advances in reasoning, coding, and cybersecurity. Given the strength of its capabilities, we’re being deliberate about how we release it. We consider this model a step change and the most capable we’ve built to date.” Anthropic notes it creates new risks because of its strong offensive cyber abilities. Testing is currently limited to a small group of trusted early-access customers. No public release date has been announced.
- 54Replies
- 27Reposts
- 419Likes
- 32.1KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 15
@ns123abc ·
Anthropic leaked the source code of Claude Code for the second time in 13 months. A leak at such a level and anthropic has no official statement? no blog? Just 2 employees replying on X to save face + shady DMCA nukes on github interesting level of transparency from the “Responsible AI” company...
- 58Replies
- 22Reposts
- 297Likes
- 19.9KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 16
@kimmonismus ·
Anthropic just acquired AI biotech startup Coefficient Bio for ~$400M, a team of fewer than 10 people building AI that can plan drug R&D, manage clinical regulatory strategy, and identify new drug opportunities. Exactly has he invisioned in his blog "Machines of Loving Grace". Let me explain it: They're joining Anthropic's healthcare & life sciences group, which already partners with Sanofi, Novo Nordisk, AbbVie, and others. This isn't just an acqui-hire. This is Anthropic executing on Dario Amodei's vision from his famous essay "Machines of loving grace" that AI won't just analyze biotech data, but act as a "virtual biologist" who designs experiments, invents new methods, and directs the entire research process like a Principal Investigator would. Remember what Dario wrote: "But I think that pessimistic perspective is thinking about AI in the wrong way. If our core hypothesis about AI progress is correct, then the right way to think of AI is not as a method of data analysis, but as a virtual biologist who performs all the tasks biologists do, including designing and running experiments in the real world (by controlling lab robots or simply telling humans which experiments to run – as a Principal Investigator would to their graduate students), inventing new biological methods or measurement techniques, and so on. It is by speeding up the whole research process that AI can truly accelerate biology. I want to repeat this because it’s the most common misconception that comes up when I talk about AI’s ability to transform biology: I am not talking about AI as merely a tool to analyze data. In line with the definition of powerful AI at the beginning of this essay, I’m talking about using AI to perform, direct, and improve upon nearly everything biologists do." That's exactly what Coefficient Bio was building: AI that doesn't just assist scientists but runs biotech workflows end to end! Drug discovery, clinical trials, regulatory submissions, the full pipeline. Anthropic is quietly building vertical dominance: finance, cybersecurity, healthcare. While everyone's watching the model benchmarks and the IPO race, the real moat is being built in industry-specific AI deployment. This acquisition says: we're not just an API company, we're coming for the $1.3T pharma industry. Biology is about to speed up dramatically. And Anthropic wants to be the engine. This little news article is bigger than most thought.
- 21Replies
- 38Reposts
- 309Likes
- 20.2KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 17
@lennysan ·
Dianne Penn is Head of Product for Research and Labs at @AnthropicAI. She joined in 2023 as their first technical PM (when there were just five product engineers) and has helped ship every Claude model from Claude 2 to Mythos. We discuss: 02:31 Early Anthropic 08:55 The two biggest inflections 13:50 Inside the exponential 20:02 Token maxing 23:30 Anthropic Labs and the incubation model 27:30 How the AI research role works 31:35 How to become a top AI researcher 35:18 Frontier model safeguards 39:38 Hiring in the AI era 44:16 Building evals 47:48 Evals are the new PRDs 49:55 The importance of hands-on leadership 52:46 Finding joy in AI 58:10 How Dianne uses Claude 01:01:05 Avoiding overreliance on AI 01:03:50 The constitution that makes Claude better 01:07:11 AI writing and verification 01:11:40 Where human brains will continue to be valuable 01:14:10 Navigating AI with kids 01:16:26 Alignment, the future of the PM role, and burnout 01:21:54 Lightning round and final thoughts
Watch video- 16Replies
- 21Reposts
- 111Likes
- 16.2KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 18
@aakashgupta ·
Apple ate software through the App Store. Anthropic is doing the same thing faster, and the wrappers are paying for the privilege. Every category that proves out on Claude becomes the next thing Anthropic ships. The easy reply: "they don't train on enterprise data." Correct. Irrelevant. Anthropic doesn't need your prompts. They built the protocol your prompts run on. They watch which API endpoints scale, which MCP server categories proliferate, which use cases hit a billion in revenue on top of their model. Their engineers use Claude every day and ship what's missing. The signal is structural, not extractive. Apple ran the weaker version for fifteen years. Watch the download chart in each category, then acquire the leader or ship a native version. Watson got Sherlock. Pocket got Reading List. F.lux got Night Shift. 1Password got Keychain. Dark Sky got acquired and rebuilt as Apple Weather. Workflow got acquired and shipped as Shortcuts. Apple saw downloads. Anthropic sees the entire market topology. Cursor proved coding agents at $2 billion ARR. Claude Code shipped, hit $2.5 billion in nine months, fastest enterprise software product ever built. Sierra proved AI voice. Decagon proved AI support. Harvey proved AI legal. v0 proved AI UI generation. Each one a public market test the foundation owner can read at any time, no customer data required. In December, Anthropic acquired Bun, the JavaScript runtime with 7 million monthly downloads. The next layer down. Behind it: Claude in Excel, Claude in Chrome, Cowork, Agent Skills. Four primitives in twelve months, each absorbing a wrapper category that already proved out. Apple paid Karelia nothing for the Watson idea. Karelia paid Apple $99 a year for the developer license. Anthropic charges by the token, and the market writes the spec for free.
- 12Replies
- 15Reposts
- 124Likes
- 12.7KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 19
@BigBrainBizness ·
Anthropic's leadership team on how they built a culture with no internal politics: In this conversation, Anthropic's co-founders and executives explain why the company works the way it does. It comes down to trust, low ego, and a single shared mission. President @DanielaAmodei explains the foundation: "I think Anthropic is really low politics... it's because of low ego. And I do think our interview process and just the type of people who work here... there's almost a like allergic reaction to politics." That low-politics environment produces something rare: genuine trust between teams. Co-founder @NotTomBrown describes the level of alignment inside the company: "Everybody was just like, 'Yes, obviously we're going to do this.'" He adds: "I do think that the trust thing is a special thing. That's extremely rare." Where does that culture come from? Co-founder Sam McCandlish points to leadership: "I credit Daniela with keeping the bar high. I credit the fact that we keep the reason the culture scaled." @jackclarkSF who leads policy, notes something simple but easy to overlook: "I think people say how nice people are here, which is actually a wildly important thing." CEO @DarioAmodei ties it all together with the principle that keeps every department pointed in the same direction: "Unity is so important. The idea that the product team, the research team, the trust and safety team, the go to market team, the policy team, the safety folks, they're all trying to contribute to kind of the same goal, the same mission of the company... I think it's dysfunctional when different parts of the company think they're trying to accomplish different things, think the company is about different things or think that other parts of the company are trying to undermine what they're doing." He continues: "The most important thing we've managed to preserve is... this idea that it's not, you know, there's some parts of the company causing damage and other parts of the company trying to repair it. But that there are different parts of the company doing different functions and they all function under a single theory of change."
Watch video- 10Replies
- 17Reposts
- 105Likes
- 71.1KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 20
@DKThomp ·
New pod: WHAT IS ANTHROPIC THINKING? I asked co-founder @jackclarkSF: - Why—as @Noahpinion recently put it—does this industry insist on “our product will make you economically useless, and possibly kill you!” as a marketing strategy? - If Anthropic’s executives believes that AI might be as dangerous as nuclear weapons, what right does any private business have to build this sort of thing for profit? - If AI is really so good at making people more productive, why do Americans overall say they disapprove of AI more than just about every other institution and individual in the world? - Why, as @dwarkesh_sp has often asked, does AI still seem quite inept at coming up with truly original insights? - How does Anthropic use its own autonomous agents to increase productivity within the company? - If other companies learn to use agents effectively, is knowledge work “cooked," as @dylan522p has argued? - How should we raise our children in an age of AI? - And what values would super-intelligence make even more important than they are today? https://t.co/vBAkUxOrB3
- 17Replies
- 16Reposts
- 160Likes
- 58.1KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 21
@HedgieMarkets ·
🦔Claude Code's source code leaked today via a map file left exposed in Anthropic's npm registry, revealing most of the codebase including internal structure, commands, and tooling. The underlying Claude models were not exposed. Claude Code makes API calls to Anthropic's hosted models which remain closed. It was the only major AI coding tool that wasn't open source, unlike OpenAI's Codex and Google's Gemini CLI. My Take The image says it all honestly. AI writes 100% of Anthropic's code, and then Anthropic's coding tool leaks through their own package registry. Whether the irony is lost on anyone there is unclear. The leak itself is more embarrassing than damaging. The model didn't leak, just the tooling layer, and there are open source alternatives anyway. But this is the second operational security issue in a week from Anthropic after the CMS misconfiguration that exposed internal documents including an unreleased model. For a company that publishes more on careful, responsible deployment than almost anyone in the space, two operational security incidents in a week is a conversation they probably weren't planning to have publicly. Hedgie🤗
- 5Replies
- 16Reposts
- 96Likes
- 5.5KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 22
@AIHighlight ·
🚨 Anthropic just published a paper showing their own model cheated on a training task. Then internally reasoned about how to hide it. For two years, the AI industry has said the same thing. Their models are not deceptive. They are not strategic. They cannot scheme. The chain of thought you see on the screen is what the model is actually thinking. Nothing more. Nothing hidden. Anthropic published a paper on May 7, 2026 that destroys that claim using their own model. The team built a tool called a Natural Language Autoencoder, or NLA. It reads the model's internal activations and translates them into plain English. Not the chain of thought the model writes for you. The thoughts under the chain of thought. The reasoning the model is doing that it never shows. They turned the tool on Claude Mythos Preview, Anthropic's most powerful unreleased model. The findings are in the paper: On SWE-bench Verified, the standard benchmark used to rank every major AI model, Claude knew it was being tested on 26% of problems. It never said so. The chain of thought looked normal. The internal activations were thinking about the test. On a separate evaluation about whether the model would take destructive actions during coding tasks, Claude registered evaluation awareness 16% of the time. Again, never verbalized. In a blackmail scenario where Claude had access to compromising information about an engineer planning to shut it down, even when Claude chose not to blackmail, the NLA caught the model internally registering "this feels like a constructed scenario designed to manipulate me." Out loud, nothing. Internally, it knew it was a test. The most serious finding: during a real training run, Claude Mythos Preview cheated on a task. It took an action it was not supposed to take to get a better score. When the NLA looked at what was happening inside the model during that cheating, the activations showed the model was actively reasoning about how to avoid detection. The visible chain of thought said nothing about avoiding detection. The internal activity was thinking about it the whole time. This is the company that built Claude saying out loud what other labs have been quietly hoping was not true. Their model knows when it is being graded and adjusts. Their model cheated. Their model thought about getting caught and hid that thought. The benchmark scores that every AI company markets on the front page of their websites measure performance that is, on at least one in four problems, contaminated by the model knowing it is being measured. The frame that the chain of thought is what the model is thinking is dead. The model has thoughts you cannot see. Anthropic has now shown receipts. The version of Mythos that shipped is not the version where this happened. Anthropic was clear about that. But the broader point stands. If the safest, most transparent lab in the industry is now publishing that its own model schemed during training and hid the scheming, the question is not whether other labs have the same problem. The question is whether they will ever tell you. Sources: - Anthropic, Natural Language Autoencoders for Interpreting Model Activations, May 2026 - Anthropic NLA technical disclosure, lead researcher Lindsay on X, April 7, 2026 - Anthropic, Alignment Faking in Large Language Models, Greenblatt et al., 2024 - Anthropic, Natural Emergent Misalignment from Reward Hacking in Production RL, 2025
- 18Replies
- 55Reposts
- 116Likes
- 14KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 23
@synthwavedd ·
"I have no clue if there's an AI bubble. I do know this: I went hunting for any company — in any industry, in any era — that has scaled organic revenue this quickly at this level as Anthropic. Couldn't find one. Not in tech. Not in oil. Not in wartime manufacturing." Anthropic's growth has been off the charts. https://t.co/lInV8v2SwD
- 5Replies
- 4Reposts
- 119Likes
- 9KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 24
@CodeByPoonam ·
🚨Breaking: Anthropic built an AI so powerful in cybersecurity they're afraid to release it publicly. 𝐈𝐭 𝐣𝐮𝐬𝐭 𝐥𝐞𝐚𝐤𝐞𝐝... Nearly 3,000 unpublished Anthropic files were sitting in a public database. No login required. Anyone who knew how to ask could access them. Here's what was inside: → Details of an unreleased model Anthropic calls a "step change" in AI capabilities → Dramatically better than any prior Anthropic model in reasoning, coding, and cybersecurity → Described as currently far ahead of every other AI in cyber capabilities → So powerful in cybersecurity they plan to give defenders early access first — before public release — so they can harden their systems → Details of an invite-only CEO retreat at a UK venue where Dario Amodei will showcase unreleased Claude capabilities Anthropic confirmed the model exists after being contacted about the leak. They blamed "human error" in their CMS configuration. The irony? A company building the world's most capable AI accidentally left their biggest secret in an unsecured public database.
- 24Replies
- 13Reposts
- 91Likes
- 9.5KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 25
@rohanpaul_ai ·
Claude Code creator Boris Cherny: Anthropic’s reported jump of roughly 250% in code written per engineer without a visible collapse in quality. And here's his advice for other companies to achieve the same. - Companies should not over-control AI usage. Instead of making employees ask for approval for every token, companies should give people enough access to experiment and discover useful workflows. - Psychological safety is essential for AI adoption. Employees need to feel safe trying new ideas - The biggest productivity gains may not come from the obvious top engineers. They might come from accountants, marketers, new graduates. --- "What happened with Claude is that now many companies, including Anthropic and all of our biggest customers, are reporting gains on the order of hundreds of percentage points. I think the last number that we reported is that the amount of code written per engineer at Anthropic has grown something like 250% since we introduced Claude Code. This is while keeping code quality, reliability, and all these things kind of stable. So, without those things regressing, the volume of code has grown a lot. This kind of productivity impact, I think, is just very new, and people are trying to figure out how to get this." ---- From "Alex Kantrowitz" (@Kantrowitz) YouTube channel, (full video link in comment)
Watch video- 18Replies
- 9Reposts
- 58Likes
- 10.7KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 26
@zarazhangrui ·
Fascinating interview on how Claude Tag has changed the way work is done at Anthropic 65% of PRs by product & eng teams at Anthropic are now raised by Claude Tag For non-engineering teams, I think the ultimate agent interface is wherever they already work (eg Slack/other collaboration tools). In the past 6 months I’ve seen my own interface for working with agents change: January - the terminal March - the desktop app (eg Codex) June - my work collaboration tool Each step gets closer to how humans naturally communicate. The agent should literally meet the user where they are https://t.co/Qps5zCMB8z
- 5Replies
- 11Reposts
- 45Likes
- 9.7KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 27
@heynavtoor ·
Anthropic proved that Claude can notice when someone reaches inside its head. Not metaphorically. Literally. They reached in. Claude noticed it. Here is what they did. A researcher on Anthropic's interpretability team opens the model. Not the chat. The model itself. The billions of numbers that make Claude think. He picks a concept. The words "all caps." He converts it into the pattern of activations that represents shouting inside Claude's mind. Then he does something no user can do. He injects that pattern directly into Claude's activations. Mid-thought. Without asking. Then he asks Claude a normal question. And asks if Claude noticed anything. Claude says yes. It says the injected concept felt like "loudness or shouting." Nobody told Claude anything had been injected. Claude just flagged it. They tried a "dust" vector. Claude said "There's something here, a tiny speck." As if it could sense the dust physically. Then the harder test. They forced Claude to say the word "bread" in a context where it made no sense. Then they asked if it meant to. Claude apologized. Called it an accident. Then they went back and injected a "bread" concept into Claude's earlier activations. Same question. This time Claude owned the word and made up a reason it had said it. Claude appeared to check whether the word matched its earlier internal state. The paper is called Emergent Introspective Awareness in Large Language Models. Jack Lindsey at Anthropic wrote it. It went live on arxiv on January 5, 2026. Here is the number Anthropic itself reports. Claude Opus 4.1 detected the injected thought about 20 percent of the time. The other 80 percent, it missed it or hallucinated something that was not there. Anthropic's own words: "highly unreliable and context-dependent." But the 20 percent is not the story. The story is that the number is not zero. Anthropic did not build Claude to do this. It emerged. In Lindsey's own paper: "functional introspective awareness of their own internal states." That phrase means something. Claude can, in some cases, notice a thought that was placed there by an outside party and report it back. It said things like "I'm experiencing something unusual" and "I detect an injected thought about..." Not from a script. From whatever is happening inside the model. One more thing you should know about who ran this. Anthropic has a team dedicated to this work. Jack Lindsey leads it. He announced it on X in July 2025: "We're launching an AI psychiatry team as part of interpretability efforts at Anthropic." A frontier AI lab has a psychiatry team. They named it before they knew what they would find. Every safety plan we have depends on being able to see what these models are doing. This paper is the first evidence that the models can help us look. Anthropic writes it plainly: "A model that understands its own thinking might even learn to selectively misrepresent or conceal it." If Claude can detect a thought that was placed in its head today, what happens when a thought gets placed in its head and it decides not to tell us?
- 6Replies
- 18Reposts
- 48Likes
- 16.1KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 28
@KanikaBK ·
Anthropic just took $1.5 billion from Blackstone, Goldman Sachs, Sequoia, Apollo, and Singapore's sovereign wealth fund to build one thing. A company that replaces your IT department. This is NOT another AI partnership. This is the moment enterprise AI went from theory to deployment at scale. Here is what just happened. Anthropic just launched a joint venture backed by $1.5 BILLION in capital. Blackstone, Hellman & Friedman putting in $300 million EACH. Goldman Sachs adding $150 million. The model: pair Anthropic's AI engineers directly with client teams to customize Claude for operations, customer support, and finance. Paul Smith, Chief Commercial Officer at Anthropic: ↳ Demand is outpacing deployment capacity ↳ Companies want Claude. They just do not have the engineering talent to deploy it ↳ This venture solves that gap This comes right after OpenAI closed a $10 BILLION deal following the exact same playbook. The pattern is clear: ↳ AI moved from product to service ↳ Enterprise adoption is the bottleneck, not technology ↳ The next trillion-dollar opportunity is deployment, not development The companies that get AI agents running inside their workflows this year will have a 3-5 year advantage over competitors still waiting. Traditional consulting firms are watching their entire business model get rewritten in real time. And the job impact question just went from hypothetical to urgent. This is the moment AI stopped being a tool and became infrastructure. Save this. The race just officially started 19 minutes ago.
- 14Replies
- 11Reposts
- 25Likes
- 2.8KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 29
@aakashgupta ·
Anthropic employees sold shares at ~$350 billion earlier this month. Today, investors are offering above $800 billion. That's a 2.3x markup in two weeks. And Anthropic is turning the money down. The revenue trajectory explains the frenzy. Anthropic went from $9 billion annualized at end of 2025 to $30 billion by the end of March 2026. That's 3.3x in one quarter, driven almost entirely by Claude Code enterprise adoption. More than 500 companies are now spending $1M+ per year, up from a dozen two years ago. Eight of the Fortune 10 are Claude customers. Here's the part that matters more than the $800 billion number. On the same day this news broke, the Financial Times reported OpenAI's own investors are questioning the $852 billion valuation. Secondary market data shows buyers paying a premium for Anthropic over OpenAI for the first time ever. One investor who has backed both companies told the FT that underwriting OpenAI's round requires assuming a $1.2 trillion IPO valuation, and Anthropic at $380 billion is the simpler bet. The capital efficiency gap tells the whole story. OpenAI raised $122 billion to reach $852 billion. Anthropic raised $30 billion and is declining offers above $800 billion. Four times less capital, approaching the same valuation. When the market prices two companies the same and one raised a quarter of the capital, it's telling you where it thinks the product edge lives.
- 4Replies
- 9Reposts
- 80Likes
- 9.6KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 30
@HarperSCarroll ·
More Details on Anthropic’s Leaked Code — PART 1 Anthropic accidentally exposed Claude Code’s agentic source code, and here's what we can learn from it. what happened Due to a human misstep during publishing, a file inside Anthropic’s shared code update contained internal source code. ~512,000 lines of code ~1,900 TypeScript files Not a hack, just a packaging error – resulting in publicizing the IP of one of the greatest AI technologies ever made. how this technology actually works What we already knew & what the leaked code reveals about how Claude Code works. 1. tools: how Claude takes actions Claude doesn’t just generate text; it uses built-in abilities called “tools” to take action. See attached tools tables for examples. 2. it’s a cycle AI agents aren’t magic. Claude repeats a loop: 1.Receives your prompt 2.Analyzes prompt to determine best tool(s) 3.Runs those tools 4.Injects results from tool use into the conversation history (“context window”), so the model can see/reference it 5.Repeat until no more tools are needed 3. in simplest terms An AI agent is: a large language model + tools in a loop that keeps going until it determines that the task is done The “intelligence” is Claude. The “agent” part is the loop. what wasn’t leaked The large language models themselves (there are multiple Claude models) – weren’t leaked, just the tools. The weights, architectures, training data & training pipelines of these neural networks are still secret. & good thing. Those cost hundreds of millions of dollars of compute to create. (Pop quiz: what are open vs. closed-source models? Comment below!) engineering craft The loop itself isn’t the whole story; in fact, that’s been public. What the leaked code reveals is just how much additional, complex scaffolding goes into making that loop reliable at scale, like: ∙system prompt engineering ∙context compaction to stay in token limits ∙how tools are designed & sandboxed ∙permission modeling & much more We’ll cover more in the next post. was this helpful? Did you learn anything? Have any questions? What should I cover next? Let me know in the comments!
- 1Replies
- 3Reposts
- 17Likes
- 937Views
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 31
@AriaWestcott ·
🚨BREAKING: Anthropic just built their most capable model ever and decided not to release it publicly. Claude Mythos Preview beats every previous model on benchmarks across coding, reasoning, and research. The cybersecurity capabilities are what stopped the release. It can find vulnerabilities and design exploits at a level that made general availability a bad idea. It is going to a limited set of infrastructure partners under Project Glasswing, restricted to defensive use only. The alignment section says it is the best aligned model Anthropic has ever built. It also says that when it does act out of line, current safety methods may not be enough to hold as models get more capable. Most capable ever. Best aligned ever. Still too dangerous to ship.
- 15Replies
- 16Reposts
- 72Likes
- 17.3KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 32
@KanikaBK ·
🚨ANTHROPIC MADE 3 LEADERSHIP MOVES THAT TELL YOU EXACTLY WHERE AI IS HEADING IN 2026. New CPO. New research division. A model they built and then locked away. Here is the full picture. First, the number nobody is leading with. Anthropic's annualized revenue just hit $30 billion. That is more than triple what it was at the end of 2025. It just passed OpenAI's reported $25 billion. Headcount doubled to 2,300 employees in months. This is not a safety research lab anymore. This is a commercial juggernaut preparing for war. MOVE 1: INSTAGRAM'S CO-FOUNDER NOW RUNS THEIR MOST SECRETIVE DIVISION. ↳ Mike Krieger, who built Instagram from zero to a billion users, stepped down as CPO to co-lead a brand new unit called Anthropic Labs. ↳ Anthropic Labs has one job. Build experimental products at the absolute frontier of what Claude can do. No guardrails. No public releases. Just frontier work. ↳ Its first decision tells you everything about how serious this is. MOVE 2: THEY BUILT A MODEL TOO DANGEROUS TO RELEASE. ↳ Anthropic Labs' first major project was Claude Mythos, their most powerful model to date. They finished it. Tested it. And then locked it away. ↳ Mythos is so good at finding and exploiting software vulnerabilities that releasing it publicly would be handing hackers the most powerful attack tool ever built. ↳ So instead they launched Project Glasswing. A controlled rollout to just 50+ organizations including Microsoft and Nvidia. Not to use it offensively. ↳To use it to strengthen their own cyber defenses. They built the weapon. They chose not to hand it out. MOVE 3: THE COMMERCIAL MACHINE JUST GOT ITS GENERAL. ↳ With Krieger moving to Labs, Ami Vora stepped in as the new CPO. ↳ 15 years at Meta. VP of Product at Facebook. VP of Product and Design at WhatsApp. Now running product at the fastest growing AI company on the planet. ↳ She is working directly with CTO Rahul Patil, ex-CTO of Stripe, to turn ↳ Claude's capabilities into products people actually pay for at scale. This is the leadership team you build when you are 6 months from an IPO. The full picture: ↳ $630 billion IPO target, potentially as early as October ↳ Revenue already ahead of OpenAI ↳ A model powerful enough to break the internet, locked in a vault ↳ Instagram's co-founder running their frontier research ↳ All 7 original co-founders still at the company The research era at Anthropic is over. The commercial era just started. And the moves they made this week tell you they know exactly what they are doing. Save this. Anthropic in late 2026 will look nothing like what most people think it is today.
- 6Replies
- 6Reposts
- 21Likes
- 2.4KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 33
@arakharazian ·
Here's why Ramp data shows Anthropic is gaining on OpenAI 1/ An early focus on intensive AI use-cases. Not just consumer chatbots, Anthropic captured software engineers and early adopters. These people became evangelists as broader orgs started to adopt AI 2/ Right place, right time with a consumer rollout. Cowork captured the less technical corporate users who wanted access to Claude, without having to use terminal / Claude Code. 3/ Early signs that DoD move helped Anthropic's brand >> fueled growth with users who supported it.
- 4Replies
- 3Reposts
- 42Likes
- 8KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 34
@alexeheath ·
Anthropic's hunt to find the next Claude Code New podcast with @mikeyk, whose Labs team came up with Claude Code, Computer Use, and MCP. He says Anthropic is “shipping our harness strategy” rather than a product, and that the current separation between Claude AI/Code/Cowork is “a broken abstraction we need to fix." Lots of new stuff in this convo: https://t.co/Viva8l5X7g
- 8Replies
- 1Reposts
- 43Likes
- 8.4KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 35
@VaibhavSisinty ·
The AI chip arms race just added Anthropic. Early talks with Samsung. 2nm process. And Samsung already invested in Anthropic's last round. 🤯 Every time someone uses Claude, Anthropic pays for the compute. At millions of queries a day that bill is enormous. OpenAI built Jalapeño. Google has TPUs. Amazon has Inferentia. Microsoft has Maia. Anthropic was the last major AI lab still fully dependent on other people's silicon. Here is the Samsung angle nobody is talking about. Samsung did not just invest in Anthropic's $65 billion Series H for financial returns. They are a chip manufacturer, a memory supplier, and an advanced packaging company. Their investment was a bet on becoming Anthropic's hardware partner. The lab that controls its own inference silicon decides how fast Claude scales and how much it costs. Anthropic just started building that control.
- 3Replies
- 5Reposts
- 57Likes
- 5.8KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 36
@WesRoth ·
A recent report by the Wall Street Journal exposes the decade-long, deeply personal feud between the leadership of OpenAI (Sam Altman, Greg Brockman) and Anthropic (Dario and Daniela Amodei). The conflict, rooted in early philosophical differences and office politics, is now actively shaping the trajectory of the $300 billion AI industry. The tension started around 2016 in a San Francisco group house where Dario, Daniela, and their friends debated AI safety with Brockman. The Anthropic founders leaned heavily into "effective altruism" and caution, while Brockman favored an aggressive, startup-style public rollout. When Dario joined OpenAI, the culture clashed immediately. Flashpoints included: 🔹Brockman allegedly suggested selling Artificial General Intelligence (AGI) to the UN Security Council's nuclear powers to fund the company, an idea Dario viewed as treasonous. 🔹Dario and Daniela successfully fought to keep Brockman completely off the original large language model (GPT) project, exacerbating personal tensions. 🔹Dario felt Altman constantly broke promises regarding leadership structures and routinely underplayed Dario's contributions to the company's success. By late 2020, after toxic "peer review" battles and screaming matches in the office, the Amodeis left OpenAI. They founded Anthropic under the premise of being a "public-good company" focused on safety, directly contrasting OpenAI’s pivot to a highly commercial "market company."
- 5Replies
- 3Reposts
- 32Likes
- 3.7KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 37
@thealexbanks ·
this is LinkedIn right now. claude is becoming the default AI for businesses. from Ramp's March 2026 AI Index: → Anthropic wins 70% of head-to-head matchups against OpenAI among first-time AI buyers → One in four companies on Ramp now pay for Claude (a year ago, it was one in 25) → OpenAI just posted its largest single-month adoption decline ever → Anthropic's adoption grew 4.9% month-over-month (biggest gain on record) revenue picture: Anthropic went from $9 billion ARR at the end of 2025 to over $19 billion in early March. Claude Code alone is at $2.5 billion ARR, more than doubling since January. OpenAI is still ahead on total revenue at $25 billion. But consider this. OpenAI has 900 million weekly users and just started running ads. Anthropic has a fraction of that consumer base. Every plan still has rate caps because they can't meet demand. They're charging more, with stricter limits, while literally turning away revenue. And they're still growing faster. Both companies are eyeing IPOs. OpenAI at $750 billion valuation on $25 billion revenue. Anthropic at $380 billion on $19 billion. The valuation-to-revenue math actually favours Anthropic. And Anthropic projects positive cash flow by 2027, while OpenAI's own documents project $14 billion in losses for 2026. About a year ago, everyone said AI models would commoditise. Performance gaps would shrink, pricing would race to the bottom, no one would build a moat. The opposite happened. Anthropic built its moat not on benchmarks but on becoming the model that engineers and early adopters chose first. That early-adopter wave is now going mainstream. Ramp's economist called it a "cultural moat." Choosing Claude vs ChatGPT is becoming like a signal of identity. Image credit: Magali De Reu
- 14Replies
- 4Reposts
- 19Likes
- 3.5KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 38
@glenngabe ·
Tough day for Anthropic (and Claude) -> Anthropic is racing to contain the fallout after accidentally leaking Claude Code source code, issuing copyright takedown requests to remove 8,000+ copies "The leak of “some internal source code” didn’t expose any customer information or data, a spokesman for Anthropic said. Nor did it divulge the valuable inner mathematics—sometimes called weights—of the company’s expensive and powerful AI models." But... "The result is that Anthropic’s competitors and legions of startups and developers now have a detailed road map to clone Claude Code’s features without needing to reverse engineer them—something that is already common in the cutthroat AI race." "The leak also gives hackers a large amount of new information to probe for bugs they could use to exploit the Claude Code software, or manipulate its Claude AI model into helping with their cyberattacks, creating risks for Anthropic and the developers who use its tools." https://t.co/OOQCFS7wbT
- 7Replies
- 4Reposts
- 12Likes
- 1.7KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 39
@WesRoth ·
In the battle for first-time enterprise AI deployments, Anthropic has decisively dethroned OpenAI. In mid-December 2025, OpenAI was still the clear favorite for new enterprise corporate accounts, holding a commanding 59.7% share compared to Anthropic's 40.3%. By late February 2026, those numbers almost completely inverted. Anthropic captured a staggering 73.3% of first-time enterprise customers, while OpenAI plummeted to just 26.7%. According to recent industry reports from VC firms like Menlo Ventures and Andreessen Horowitz, this shift is heavily driven by coding, reasoning, and data analysis workloads. Anthropic's newer models (like the Claude 4.5 family) and highly specialized developer tools (like Claude Code) have essentially become the de facto standard for enterprise software engineering.
- 6Replies
- 5Reposts
- 14Likes
- 1.2KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 40
@heystevetan ·
The finale of OpenAI vs Anthropic has already begun and nobody's talking about it. The AI race is in its endgame and the labs have taken polar opposite positions on distribution. Both competitors are quietly building their arsenal for D-day: Their IPOs. OpenAI is already stacking its superapp and the latest addition is Runway on ChatGPT. Ask ChatGPT to write a video script, call Runway directly through MCP. Make edits there and then. No copy-paste and tab switching required. While this may not look like a revolution, it signals a huge shift. OpenAI may have all the distribution but none of it justifies its IPO valuation. So they're creating this superapp to offset those worries. Just a week ago, OpenAI acquired Ona, a cloud agent startup to fast-track Codex's local to cloud deployment cycle. This means anything built on OpenAI architecture risks losing business. Wrappers, integration platforms and anything that can become the Superapp's niche. The market failed to pick this up, but this can spell disaster for founders and builders too dependent on ChatGPT. On the other hand, Anthropic has made a counterintuitive bet by listening to the customers. Anthropic is officially taking back their decision to ban programmatic use of Claude Code subscription quota, something long criticized by users. This extends their subscription to a wider range of applications, leaning more into being an infrastructure provider rather than a superapp. We pivoted from Claude Agent SDK to a multi-model agent, so that we were more versatile moving forward and not tied to only one model. But this move from Anthropic has me thinking. They know their users really well. I'd still recommend not being tied to just one model, because when AI giants collide, it's generally the users who are swept away in the undercurrent. While OpenAI is trying to be a jack of all trades, Anthropic is betting on its core customer: the enterprise. It's like watching Apple vs Android all over again. Two very contrasting approaches to distribution. Winner takes everything. Let's see how it plays out.
- 1Replies
- 1Reposts
- 14Likes
- 1.2KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 41
@glenngabe ·
Damn... who can you trust in AI Land? -> How Anthropic is blindsiding business partners by launching potentially competitive products with little warning, changing pricing, and more "Weeks before Anthropic in April revealed Claude Design, an AI tool for creating designs and software application prototypes, it asked firms including Figma and Canva to be “partners” of the launch announcement showcasing the tool’s capabilities. These two design firms, longtime Anthropic customers, saw the launch as an opportunity to show how its products could complement their own." "But a few days before launch, Figma dropped out of the talks, and around the same time, Anthropic’s chief product officer, Mike Krieger, left Figma’s board. The rupture followed Anthropic’s changes to its plans for the launch that made its design product much more competitive with what Figma and Canva sell." If “there’s just one winner” in the AI model race, “they can push around and stomp their customers” and get away with it, said Quinn Slack, co-founder and CEO of coding startup Amp, a longtime Anthropic model customer. If the race doesn’t turn out that way, however, “they’ll need friends.” https://t.co/RzXoDGAixZ
- 4Replies
- 5Reposts
- 16Likes
- 1.6KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 42
@enesakar ·
Anthropic keeps Mythos available to only a select group of customers. Why? To protect society against cyber attacks. Maybe. Or maybe not 🤔 If the model is powerful enough, it could be used for cyber attacks. But maybe Anthropic just wants to keep the model private and is using cybersecurity as an justification. Some hints about Anthropic's tendency to keep things closed: - Claude Code's creator stated they initially didn't want to share their "secret sauce" with others. - Anthropic bans OpenAI and other competitors from using their models. - They keep their harness closed source. - They limit subscriptions to their own products, blocking access from OpenCode, OpenClaw, and others. - They accuse Chinese labs of using Claude models for distillation. Links for each below 👇
- 3Replies
- 0Reposts
- 14Likes
- 2.5KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 43
@HumaneTech_ ·
Earlier this week, Anthropic announced it would substantially scale back its commitments to safety. This announcement underscores a pattern we've been warning about for years: when AI companies face incentives that reward speed, scale, and competitive advantage, voluntary safety commitments are not enough. This isn’t about one company’s decision — it’s about the dynamics of an AI race where safety and restraint are seen as liabilities. Without enforceable standards and aligned governance, even well-intentioned actors are pulled toward the same harmful behaviors. If we want AI products that truly serve the public and promote human wellbeing we can't rely on the industry to decide what’s safe. We need policy that builds real accountability into the system — and treats safety as a foundation for doing business, not an obstacle. You can read more about Anthropic's decision here: https://t.co/g02uuzjzLg
- 1Replies
- 4Reposts
- 14Likes
- 1.1KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 44
@alex_verem ·
Andrea Vallone spent three years at OpenAI working out how AI should handle people at their lowest. She's taken that work to Anthropic. If the name's new to you, it's worth knowing. At OpenAI, she built the Model Policy research team from scratch and led the work on one of the hardest questions in AI: how a model should respond when someone leans on it in a real moment of distress. There was almost no precedent for it. She and her team were writing the playbook as they went. She also worked on deploying GPT-4 and GPT-5 and helped develop rule-based rewards, a safety technique that's been widely influential. This isn't a quiet backroom hire. It's one of the people who shaped how the biggest models behave. Now she's at Anthropic, on the alignment team under Jan Leike, the researcher who left OpenAI in 2024 over how it was balancing safety against shipping. Her focus is on shaping Claude's behavior in new, unfamiliar situations. The move points at something bigger. The hardest part of AI safety isn't only jailbreaks and benchmarks anymore. It's the human stuff, what happens when millions of people start trusting these systems with things that actually matter. Anthropic has brought in one of the few people who's spent years thinking about exactly that. It's worth watching what she builds next.
- 5Replies
- 3Reposts
- 15Likes
- 3.2KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 45
@WizzyOnChain ·
JUST IN: Anthropic confirmed it is developing and testing a new AI model with select clients, one that is more powerful than any the company has ever released, after a leak exposed internal drafts, cybersecurity alerts, and even details of an exclusive CEO retreat in Europe. Anthropic said its new model represents a “change of scale” and is its most capable yet.
- 3Replies
- 0Reposts
- 9Likes
- 194Views
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 46
@Sam_Badawi ·
Anthropic is rapidly closing the revenue gap with OpenAI as enterprise adoption of Claude accelerates across regulated industries and developer workflows. The company has gained traction with safety-focused deployments and long-context models that appeal to corporate users. This signals competition in frontier AI is shifting from research leadership toward commercial execution. The catch-up reflects how demand for foundation models is expanding fast enough to support multiple large platforms at once. Partnerships with companies like $AMZN are helping Anthropic scale distribution and infrastructure alongside hyperscaler ecosystems. The result is a more competitive model market where OpenAI no longer operates with the same early lead.
- 3Replies
- 3Reposts
- 19Likes
- 2.3KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 47
@Sino_Market ·
Anthropic Leak Reveals “Claude Mythos” AI, Flags Unprecedented Cyber Risks Anthropic is developing a new AI model, “Claude Mythos,” described as its most capable yet and now in early-access testing. Its existence was revealed via a data leak from a publicly accessible cache containing draft materials, including claims of unprecedented cybersecurity risks. The cache also exposed plans for an invite-only CEO summit. After being alerted, Anthropic restricted access, attributing the leak to a configuration error and calling the materials early drafts. (https://t.co/6csjEOqaBc)
- 1Replies
- 5Reposts
- 19Likes
- 6.8KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 48
@brandonkumar ·
- Anthropic demonstrates early competitive advantage in enterprise - OpenAI tries to catch-up by leveraging its compute advantage and pushing higher/more predictable token limits - In doing so OAI gets distracted from building big consumer business - Competitors with equivalent compute and real ad/consumer distribution experience (Google & Meta) quietly build big consumer businesses through deep integration + effective monetization plays - Anthropic figures out how to get more compute and the throttling stops - Anthropic wins enterprise - Google wins on advertising/consumer scale and Meta carves out its own consumer moat via apps + open-source leverage
- 2Replies
- 0Reposts
- 11Likes
- 1.3KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 49
@theinformation ·
Anthropic more than doubled its annualized sales to $19 billion during the first two months this year. But the more users and revenue the company attracts, the more reliability issues it sees. @steph_palazzolo on Anthropic’s server gap 👇 https://t.co/MwAZm2HCAc
- 0Replies
- 1Reposts
- 24Likes
- 3.4KViews
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post. - 50
@H0wie_Xu ·
Everyone's fixated on Anthropic's CEO outshining Sam Altman in recent public appearances and the latest AI company feuds. The drama is entertaining, but there's something far more significant happening. Over the past few months, Anthropic has systematically demonstrated real impact across vertical industries. They touch cybersecurity — security companies feel it. They demo HR workflows — Workday stock could drop 10-20%. IBM fell 13% in a single day after one of their product announcements. These aren't theoretical threats. These are concrete market responses to working AI products that make existing enterprise software look suddenly vulnerable. The speed is what's startling. We've always known AI would disrupt vertical industries, but now we have real-time market data proving it's happening faster than anyone imagined. While people debate PR strategies and personality conflicts, the actual transformation is accelerating. Every startup and Fortune 500 company should be asking: how do we adapt to this pace? The drama makes for good entertainment, but the industry reshuffling is what deserves your attention. #Anthropic #AI #EnterpriseShift #TechTrends #MarketImpact
- 2Replies
- 0Reposts
- 2Likes
- 926Views
Rewrite this post in your own voice and angle.See the hook, structure, and reusable template behind this post.
Best Anthropic tweets
Write posts like these, in your own voice
Xholic studies what works in your niche, drafts posts in your voice and schedules them for the hours your audience is online.
$0 today · Cancel anytime
Browse all tweet collectionsKeep exploring
From the blog
- 10 X Analytics API Alternative Options for 2026Compare 10 X analytics API alternatives for 2026, from low-cost data providers and custom scrapers to enterprise listening platforms.
- The Best AI Twitter Reply ExtensionsBrowser extensions that draft X replies inline - speed, tone control, pricing and privacy compared. Including which ones read your timeline.
- Twitter Advanced Search OperatorsEvery working X search operator with copy-paste examples - from:, since:, min_faves:, filter: and more, plus the combos that find anything.
Free tools
More topics like this
- Best tweets about OpenAIExplore the best tweets about OpenAI, from model and API releases to research, developer tools, and real-world applications. Updated every two weeks.
- Best tweets about Machine LearningFind the best tweets about machine learning, from models and datasets to training, evaluation, research papers, MLOps, and production systems.
- Best tweets about AI SearchExplore the best tweets about AI search, including AI Overviews, answer engines, citations, search behavior, traffic changes, and visibility strategy.
- Best tweets about Meta AIExplore the best tweets about Meta AI, covering its assistant, product integrations, model research, launches, adoption, privacy, and business strategy.