Competition, distribution, and moats
OpenAI’s positioning against Anthropic and open models, consumer distribution, product bundling, ecosystem integration, pricing, and defensibility for startups.
44%
Best tweets about OpenAI
50 Best Tweets About OpenAI (2026)
Explore the best tweets about OpenAI, from model and API releases to research, developer tools, and real-world applications. Updated weekly.
Substantive OpenAI product, research, and developer discussions rather than generic mentions of artificial intelligence.
- Creators
- 42
- Updated
Original Xholic analysis
What 50 top OpenAI posts reveal
Discussion in this 50-post OpenAI corpus most often concerns competition, distribution, and moats (44% of posts), followed by Codex and agentic work (34%). Posts also discuss developer tooling, ChatGPT product direction, enterprise adoption, research, and governance.
- Dominant tone
- Positive
- Median score
- 13.8
- Leading format
- Announcement
- Recent posts
- 44%
64% of posts
All-time engagement
100% of posts
Published in 90 days
Conversation map
The themes creators return to
Codex and agentic work
Codex as a coding and general work agent, computer use, long-running task execution, internal adoption, developer workflows, and agent-powered productivity.
34%
API and developer platform
OpenAI APIs, SDKs, Cookbook examples, agent frameworks, prompting, tool use, MCP, enterprise controls, evaluation patterns, and application-building practices.
24%
ChatGPT product ecosystem
ChatGPT features and product direction, including voice, work-focused experiences, browser and desktop integration, connectors, multimodal creation, and the unified app vision.
20%
Enterprise strategy and adoption
OpenAI’s business-market expansion, professional workflows, enterprise deployment, security and administration, sales strategy, and measured ROI.
20%
Models, research, and AGI
GPT model capabilities, reasoning advances, automated AI researchers, foundational research strategy, AGI definitions, and scientific applications.
20%
Agent architecture and orchestration
How OpenAI and developers design agents: tools, context, handoffs, tracing, data-agent systems, orchestration layers, and reliable domain workflows.
12%
Governance and corporate power
OpenAI’s leadership, board history, nonprofit-to-for-profit transition, IPO trajectory, institutional power, and political or public-interest implications.
12%
Tone and stance
Sentiment Positive leads
Author posture Supportive leads
Performance benchmark
- Median likes
- 56
- Median reposts
- 7
- Median replies
- 9
- Median views
- 5.7K
Posts with media make up 60% of this collection. Their median all-time score is 25.9, compared with 9.74 for text-only posts.
Format mix
- Announcement 100% · score 13.8
Consensus and debate
Where creators agree, and where they do not
Shared view
Codex is presented as extending beyond coding
Posts describe Codex being used or positioned for longer-running and cross-functional work, including engineering, QA, security, data analysis, internal tools, and product tasks. OpenAI’s own post says internal users are applying it across departments to more complex and longer-running work.
Shared view
Developer resources cover agent-building workflows
Posts highlight the Agents SDK’s agents, handoffs, and tracing primitives; the Cookbook’s code examples; and enterprise features for private MCP connectivity, identity federation, administration, and cost controls.
Shared view
A unified ChatGPT-Codex product direction is a recurring topic
Multiple posts report or characterize OpenAI as bringing ChatGPT, Codex, browsing, and desktop workflows closer together. These are reports and commentary on product direction, rather than confirmation that every described integration has shipped.
Shared view
Research and automated research are prominent in the discussion
OpenAI leadership posts emphasize foundational research and an automated scientist, while other posts discuss an automated AI researcher and safety/alignment disclosures. These reflect stated ambitions and commentary, not demonstrated future outcomes.
Open debate
Distribution advantages versus model portability
Some commentary argues that ChatGPT subscriptions, bundled workflows, and integration can strengthen Codex distribution. Other commentary argues that agent harnesses and protocol compatibility can let users substitute open models. These are competing creator interpretations, not measured market outcomes.
Open debate
Consumer-oriented ecosystem versus enterprise expansion
One post frames OpenAI primarily through a consumer personal-agent vision, while reporting and commentary also describe a stronger professional-work and enterprise push. The evidence supports both as discussed directions rather than a settled strategic classification.
Open debate
Governance discussion mixes criticism with OpenAI’s stated public-interest goals
A widely engaged post summarizes allegations from a New Yorker investigation, and other posts criticize OpenAI’s corporate evolution. In contrast, OpenAI-related posts describe goals of broad AGI access and welcome publication of safety and alignment issues. These posts present conflicting claims and perspectives.
What performs
Patterns behind standout posts
Five posts were large engagement outliers
The deterministic analytics identify five outliers: a post summarizing governance allegations (score 4023.44), a Postgres-scaling post (1912.63), a prompting-guidance comparison (1532.45), an Agents SDK post (1279.74), and OpenAI’s internal Codex-use post (520.68). The corpus median all-time score was 13.85.
Posts with media had a higher median score
Deterministic analytics report a median all-time score of 25.93 for the 30 posts with media, versus 9.74 for text posts. This describes an association in this corpus and does not establish that media caused stronger performance.
ChatGPT ecosystem was the highest-scoring theme by median
Among the supplied themes, ChatGPT product ecosystem had the highest median all-time score (36.38), ahead of governance and corporate power (30.92) and models, research, and AGI (28.31). It represented 10 posts, or 20% of the corpus.
Statistical standouts
- View standout post 1 Score 4023.4 · 290.5× median
- View standout post 2 Score 1912.6 · 138.1× median
- View standout post 3 Score 1532.5 · 110.65× median
- View standout post 4 Score 1279.7 · 92.4× median
- View standout post 5 Score 520.7 · 37.59× median
Creator landscape
Who shapes this conversation
The five most represented creators account for 20% of the selected posts.
-
1. Vaishnavi
@_vmlops
2 posts
-
2. Chris
@ChrisGPT
2 posts
-
3. GEOFF 🧠💸
@geoffwoo
2 posts
-
4. Chubby♨️
@kimmonismus
2 posts
-
5. Mark Kretschmann
@mark_k
2 posts
-
6. Rohan Paul
@rohanpaul_ai
2 posts
Vaishnavi covered OpenAI developer resources
Vaishnavi was listed among top voices with two posts and a median all-time score of 640.58. The cited posts explain the Agents SDK and the OpenAI Cookbook.
Product commentators discussed consolidation and work use cases
Chubby’s posts discuss an AI-superapp framing and the reported merging of ChatGPT, Codex, and Atlas; Mark Kretschmann’s post summarizes Codex use cases across engineering and other work. Together, these posts illustrate commentary around broader work-platform positioning.
Competition commentary emphasized distribution and workflow integration
Creators discussed Codex’s professional brand, subscription and distribution advantages, and use inside competing developer workflows. These are creator analyses rather than verified measures of competitive advantage.
Since the previous snapshot
What changed since Aug 18, 2026
- 78% of the selected posts remained.
- The creator count changed by 0.
- 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 OpenAI tweets
Top OpenAI tweets from 42 creators
Ranked 01–50
- 01
@ohryansbelt ·
The New Yorker just dropped a massive investigation into Sam Altman, based on over 100 interviews, the previously undisclosed "Ilya Memos," and Dario Amodei's 200+ pages of private notes. It's the most detailed account yet of the pattern of behavior that led to Sam's firing and rapid reinstatement at OpenAI. Here's the breakdown: > Ilya compiled ~70 pages of Slack messages, HR documents, and photos taken on personal phones to avoid detection on company devices. He sent them to board members as disappearing messages. The first memo begins with a list headed "Sam exhibits a consistent pattern of . . ." The first item is "Lying." > Dario kept detailed private notes for years under the heading "My Experience with OpenAI" (subheading: "Private: Do Not Share"), totaling 200+ pages. His conclusion: "The problem with OpenAI is Sam himself." > Sam reportedly told Mira his allies were "going all out" and "finding bad things" to damage her reputation after the firing. Thrive put its planned $86B investment on hold and implied it would only close if Sam returned, giving employees financial incentive to back him. > Sam texted Satya Nadella directly to propose the new board composition: "bret, larry summers, adam as the board and me as ceo and then bret handles the investigation." The two new members selected to oversee an independent inquiry into Sam were chosen after close conversations with Sam himself. > Before OpenAI, senior employees at Loopt asked the board to fire Sam as CEO on two separate occasions over concerns about leadership and transparency. At Y Combinator, partners complained to Paul Graham about Sam's behavior, and Graham privately told colleagues "Sam had been lying to us all the time." > OpenAI's superalignment team was promised 20% of the company's compute. Four people who worked on or with the team said actual resources were 1-2%, mostly on the oldest cluster with the worst chips. The team was dissolved without completing its mission. > Sam told the board that safety features in GPT-4 had been approved by a safety panel. Helen Toner requested documentation and found the most controversial features had not been approved. Sam also never mentioned to the board that Microsoft released an early ChatGPT version in India without completing a required safety review. > Sam made a secret pact with Greg and Ilya where he agreed to resign if they both deemed it necessary, essentially appointing his own shadow board. The actual board was alarmed when they learned about it. > Sam struck a deal with Greg to become CEO while simultaneously telling researchers that Greg's authority would be diminished, and telling Greg something different. > A board member described Sam as having "two traits almost never seen in the same person: a strong desire to please people in any given interaction, and almost a sociopathic lack of concern for the consequences of deceiving someone." Multiple sources independently used the word "sociopathic." > OpenAI is reportedly preparing for an IPO at a potential $1 trillion valuation while securing government contracts spanning immigration enforcement, domestic surveillance, and autonomous weaponry in war zones.
Watch video
- 02
@arpit_bhayani ·
Fun fact: OpenAI handles 800 million users on ChatGPT with just one PostgreSQL primary and 50 read replicas 🤯 Today, OpenAI published an engineering blog explaining how they scaled their Postgres setup to support a massive 800 million users using a single primary and 50 multi-region replicas. They dive into details around their scaling approach, the PgBouncer proxy, cache locking, and cascading read replicas. It is genuinely neat and impressive. I just published a video on my YouTube channel where I dissect the blog and break down the nuances. Give it a watch - it is short and fun.
- 03
@alex_verem ·
Both OpenAI and Anthropic just released official prompting guides. Both say the same thing. Your old prompts don’t work anymore. But for opposite reasons. Claude Opus 4.7 stopped guessing what you meant. It does exactly what you type. Nothing more, nothing less. Vague instructions that worked on 4.6? They now produce narrow, literal, sometimes worse results. Not because the model got dumber. Because it stopped compensating for sloppy thinking. GPT-5.5 went the other direction. OpenAI’s guide literally says: “Don’t carry over instructions from older prompt stacks.” Legacy prompts over-specify the process because older models needed hand-holding. GPT-5.5 doesn’t. That extra detail now creates noise and produces mechanical output. Claude got more literal. GPT got more autonomous. Both now punish the same thing: prompts written without clear thinking behind them. One developer on Reddit captured it perfectly after analyzing hundreds of community posts. The complaints tracked almost perfectly with prompt specificity. Precise prompts got better results on 4.7. Vague prompts got worse. The model didn’t regress. The prompts did. OpenAI’s new framework is “outcome-first prompting.” Describe what good looks like. Define success criteria. Set constraints. Then get out of the way. The model picks the path. Anthropic’s framework is the inverse: be surgically specific about what you want, because the model won’t fill in your blanks anymore. Two different architectures. Two different philosophies. One identical conclusion: the person writing the prompt is now the bottleneck, not the model. Boris Cherny, the engineer who built Claude Code, posted on launch day that even he needed a few days to adjust. That post got 936 likes. Meanwhile, Anthropic increased rate limits for all subscribers because the new tokenizer uses up to 35% more tokens on the same input. The model is more expensive to run lazily. Cheaper to run precisely. The models are converging in capability. The gap between good and bad output is no longer about which model you pick. It’s about the 2 minutes of structured thinking you do before you type anything. That thinking system is the skill. The prompt is just what it produces.
- 04
@_vmlops ·
OPENAI JUST OPEN-SOURCED THEIR AGENTS SDK & it's actually clean most agent frameworks are bloated... this one isn't just 3 core primitives: → agents (llm + tools + guardrails) → handoffs (route between agents) → tracing (debug every run) works with 100+ llms, not just openai. built-in session memory with sqlite or redis no manual conversation history juggling the hello world is 4 lines the multi-agent handoff is 20 18.9k stars already https://t.co/YTd1Sw44NQ
- 05
@OpenAI ·
Work at OpenAI is being transformed by agents, in every department. Across our entire company, people are using Codex to do work that is more complex, longer-running, and increasingly cross-functional. Our internal usage offers an early look at how agentic tools may reshape work as they become more capable and broadly available.
- 06
@alexxubyte ·
How OpenAI Built Its Data Agent Most teams building data agents stack routers, fine-tunes, and complex retrieval pipelines on top of multiple LLMs. OpenAI didn't. Their data agent runs on a single model and only 13 tools, across 1.5 exabytes and 90,000 tables. It's "pretty vanilla" by design. We spoke with Emma Tang, Head of Data Platform Engineering at OpenAI, to better understand the architecture and the engineering decisions behind it. The article covers: - The architecture behind the data agent - The six layers of context that make a single LLM reliable across 90,000 tables - How OpenAI Uses Codex Internally: 3 Use Cases - Five practical lessons for any team building a domain agent - Where OpenAI's data platform is headed next
- 07
@mark_k ·
OpenAI just published a new Codex use-case page, and it’s basically a catalog of what teams are already handing over to coding agents: engineering work, product work, QA, security, data analysis, internal tools, and even life-sciences workflows. Some of the coolest examples: ⬩ Reviewing GitHub PRs and understanding large codebases ⬩ Turning screenshots or visual references into responsive UI ⬩ QA-testing apps by clicking through real user flows ⬩ Refactoring legacy code, running migrations, and fixing vulnerability backlogs ⬩ Drafting PRDs, analyzing datasets, building internal apps, and assisting life-sciences research This is what coding agents look like when they stop being a demo and start becoming part of daily work.
- 08
@kimmonismus ·
OpenAI is merging ChatGPT, Codex and its Atlas browser into one desktop app and recasting Codex from a coding tool into a productivity app it says anyone can use. The figures it has been handing out to support that: 5 million weekly Codex users, enterprise revenue up 50% week over week, usage growing 5% a day. Those come from an all-hands and an internal staff note, relayed by people familiar with the remarks. Codex is increasingly evolving into a true work platform. And GPT-5.6 is also on the horizon. Great things are expected from OpenAI in the near future. Via the information
- 09
@kimmonismus ·
OpenAI "AI Superapp" announced: OpenAI says the future is not a collection of AI tools, but a single AI superapp where ChatGPT, Codex, browsing, and other agentic systems work as one. Behind that product vision is a much bigger ambition: to turn consumer scale into enterprise dominance and position itself as core infrastructure for the age of AI.
- 10
@StartupArchive_ ·
Sam Altman on the Paul Graham advice that saved Open AI: “Always make an API” Four years into OpenAI, Sam Altman and the team realized that they would have to build a really big company to fund the development of their increasingly capital-intensive foundation models. “We had this model called GPT-3,” Sam recalls. “I was turning up the urgency on the company to try and figure out a product, and we just couldn’t. It was cool, but it wasn’t good enough to make something that worked.” Then Sam remembered a piece of advice from Y Combinator founder Paul Graham that stuck with him: “You should always make an API. No matter what, you should make an API. Good stuff will happen.” Out of ideas for a product, the OpenAI team decided to make GPT-3 available as an API. “Maybe somebody will figure out something to do with it,” Sam thought. A few copywriting applications like Jasper and Copy AI did take off using the GPT-3 API, but OpenAI also noticed interesting behavior that eventually became a sleeper hit: “Some people — not a lot — would just chat with that thing all day,” Sam explains. “It wasn’t very good but there was clear user signal that people wanted to talk to the models. And given that that was the only thing besides copywriting that had real traction, we said, ‘Maybe this is just he product we should build.’” On November 30, 2022, ChatGPT was released to the public as a “research preview” using a model from the GPT-3.5 series. It reached over a million users in five days. Source: @khoslaventures (Sep 2025)
Watch video
- 11
- 12
@markchen90 ·
How does OpenAI balance long-term research bets with product-forward research fundamentals? I’ve been getting this question a lot lately, usually framed as a suggestion that Jakub (@merettm) and I are pushing an increasingly product-focused agenda. That characterization is simply wrong. Foundational research has been core to OpenAI from the start, and today we run a research program with hundreds of exploratory projects - much like the ones that led to our reasoning-model breakthrough. The majority of our compute is allocated to foundational research and exploration - and not product milestones. Anyone who has spent time with me or Jakub knows we are the last people in the world who would push for the advancement of products over the advancement of research. We’re in the business of creating an automated scientist, and capabilities that were considered grand challenges just a few years ago (like IMO-level mathematical reasoning) now emerge as normal parts of the research process. We’re also seeing our models accelerate researchers worldwide, helping advance work across biology, mathematics, physics, and even our own research. Jakub and I put a lot of effort into ensuring that research stays focused on uncovering algorithms that will scale to the compute we’ll have a year from now. We protect mindshare and amplify discourse on exploratory work. We do this while recognizing that we’re also a deployment company - and that deployment gives us access to even larger-scale compute, richer feedback, and more room for exploration. Our researchers are passionate about having their work out in the world, and a special slice of our org is dedicated to making sure our deployments are delightful for end users. Our goal isn’t to turn research into a quarterly race. It’s to build a durable research engine - one that compounds learning over time and consistently turns long-horizon exploration into real, measurable advances, while ensuring those advances become valuable in the real world. That’s the roadmap we’re executing on. And while there have been ups and downs over the last decade (as you expect with any research program), I think most of our researchers would share my strong optimism today.
- 13
@mark_k ·
OpenAI just published a major new plan for its next phase: AGI should not be controlled by a tiny number of companies, governments or individuals. The goal is to make advanced AI abundant, affordable, safe and useful enough that everyone can benefit from it. The most important detail: @OpenAI says it is building an automated AI researcher, and internally believes that by March 2028, a significant fraction of its own research may be done by AI systems working alongside human researchers. The three big goals are clear: automate more of AI research, accelerate science and economic growth, and eventually give everyone on Earth a personal AGI. This is the real post-AGI question: not just who builds the most powerful model, but who gets access to that power.
- 14
@signulll ·
okay, enough openai criticism for a second (still gonna give product feedback when it’s deserved). it’s been a little too easy. i’ve been using gpt live a lot, like a lot lot, & it’s genuinely fucking amazing. the feature to let the model speak like a human “hmm”, “let me check”, brief pauses, & natural turn taking while calling a bigger model or searching or using tools in the background is a fantastic execution. almost all of my conversations have been nothing but pleasant, i can’t describe the feeling but it’s i think a little too good at this point. nothing else is particularly close in this category right now at all. anthropic’s voice experience still feels very very underbaked (after thought really), & gemini’s personality & intelligence seem lacking. & since it’s already reaching an enormous user base, every great gpt live interaction makes people more confident in ai as a product category instead of walking away thinking ai is a gimmick. really great work.
- 15
@heynavtoor ·
🚨OpenAI was founded in 2015 as a nonprofit. Mission: "Benefit all of humanity." Sam Altman told Bloomberg in 2023: "I have no equity. I have enough money." Then Bloomberg reported OpenAI discussed giving him a 7% stake, worth $10 billion+. The board tried to stop him. He overthrew the board in 5 days. Ilya quit. Mira quit. Schulman quit. The nonprofit that built GPT with tax-exempt donations is now an $852 billion for-profit corporation. The charity gets 26%. Investors get the rest. Here's every document:
- 16
@ChrisGPT ·
OpenAI had a sharp rise in revenue in Q4, and I think they could drive a similar derivative of growth this summer by advertising GPT 6 more aggressively to the public. Now, with OpenAI pushing toward the super app vision, where everything is unified in one place with more agentic features for mobile as well, the upside can be akin to the growth they saw in Oct-Dec for revenue. If they ship a true white collar agent that can operate on your desktop and handle real tasks, plus something closer to Anthropic’s Record a workflow feature on chrome - but also adding it to the desktop (where the model can observe the screen, interact with the desktop, and learn your personal workflows), that would pull in a lot more paying subscribers. A lot of white collar people such as myself would go from just using it at work to also using it heavily for personal projects at home.
- 17
@goyalshaliniuk ·
OpenAI is making its enterprise platform more secure and easier to manage for teams building with AI. Private MCP servers now allow companies to keep their servers inside their own network while products like ChatGPT, Codex, and the Responses API connect through outbound-only HTTPS. This gives teams a safer way to connect internal tools and systems without exposing them publicly. OpenAI also introduced Workload Identity Federation and expanded Admin API controls. This means teams can manage access through cloud IAM workflows, reduce dependency on permanent API keys, set spend alerts, control model access, manage data retention, configure hosted tools, and track costs more granularly across features like file search and web search.
Watch video
- 18
@BoringBiz_ ·
Chamath on OpenAI vs Anthropic's business model People consistently try to pit these two companies against each other. But it is important to remember that the core customer base for both companies are fundamentally different OpenAI remains the consumer first model of choice for AI, while Anthropic has become deeply embedded in the enterprise motion Just one of these have a large enough TAM to justify hundreds of billions in market capitalization if you are the leading LLM provider The revenue recognition also makes it apples to oranges. OpenAI reports their revenues on a net basis, excluding Microsoft licensing fees and partnership revenue share Anthropic talks about their revenue from a gross basis, similar to ARR or bookings for B2B software companies. API fees related to usage from AWS, for example, are not excluded in the Anthropic numbers that you hear about online Ultimately, two very different revenue and business models that often end up getting compared to each other because both are leading LLM providers. Worth a listen
Watch video
- 19
@Alfred_Lin ·
At @sequoia’s AI Ascent last week, @gdb told me something that stuck: in late 2024, AI wrote ~20% of @OpenAI's code. That number is now 80%. We also got into why human attention, not compute, is the real bottleneck in AI-augmented work, plus what it might mean to run an org of 100,000 agents. We’re grateful to Greg for joining us and many of the top founders/builders in AI. You can watch the full video here: 00:00 Intro 00:49 Compute Hunger Explained 02:13 Scaling Laws Mystery 03:31 New Architectures Ahead 04:42 How Close to AGI 06:46 Startup Playbook for AI 09:24 Inside OpenAI with Codex 11:11 Teams and Governance Shift 14:52 Security and Responsible Deployment 25:33 Science Frontiers and Wrap Up
Watch video
- 20
@signulll ·
it is very interesting that openai has to build & grow another brand now (codex) that’s more professional & tied to work as opposed to chatgpt which is very much consumer land. since codex is intended for work & enterprise (which requires sales) it may not be an issue since it doesn’t have to grow organically as much. i’m curious if they ever combine the two in some way. obviously codex code is likely the basis for the new chatgpt app(s) / 90% shared code or whatever. anthropic does have a much cleaner story here although not as large of a consumer base.
- 21
- 22
@VaibhavSisinty ·
Let me explain what OpenAI just did with the new Codex update. Because most people are going to miss the actual story here. Codex is no longer a coding agent. It's a full operating layer. In one update they shipped: computer use across any app on your machine, a built-in browser with a commenting layer, native image generation through gpt-image-1.5, persistent memory across tasks, and 111 new plugins that combine skills, MCP servers, and app integrations. I've been saying for months that tools like Codex and Claude Code are becoming the default interface for everything. Not just code. Everything. Here's what most people are missing. The biggest unlock of the last year wasn't a new model. It was the shift to "do the work first by writing code, with tool calls, then talk about it." Research, ad ops, content pipelines, analytics, internal tools. All of it runs better through coding agents that call tools than through any chat interface. The people still using AI as a chat box are operating one stack behind. And now zoom out. OpenAI is reportedly shipping Spud in the next few weeks. A new base model that Greg Brockman has publicly called "two years of research" with a "big model feel." Sam Altman told employees it could "really accelerate the economy." If that lands as expected, it's at the same tier as Anthropic's upcoming Mythos. Now bolt that model onto the Codex super app I just described. Computer use. Browser. Image gen. Memory. Plugins. Skills. All running on a model designed for long-horizon agentic work. Here's the part nobody wants to say out loud. Anthropic could lose its edge of coding revenue very soon onc Spud hits Codex. And it has nothing to do with model quality. When a better coding model exists AND it's one click away from the ChatGPT tab people already pay $20-200 a month for, nobody runs evaluations. They switch that afternoon. The switching cost between Claude Code and Codex for most dev teams is approximately one API key change. No procurement. No 6-month contract cycle. You just swap. Distribution is the moat you can't build a defense against fast enough. The next era of work isn't about which AI you talk to. It's about which AI you let run your computer.
Watch video
- 23
@alexabelonix ·
Best GitHub repos to build real AI Skills in 2026: 1. OpenAI Cookbook https://t.co/4YJBN1UkoM 2. OpenAI Agents SDK https://t.co/ucBo0e5bRn 3. OpenAI Evals https://t.co/ul1vsGMBiG 4. PydanticAI https://t.co/Nhp09jJSu2 5. Hugging Face Agents Course https://t.co/tWtBM0C5jN 6. AI for Beginners https://t.co/WKZzpuhwMZ 7. Hugging Face 101 Course https://t.co/f4blxHgBhN 8. Hugging Face Smol Course https://t.co/ovCe7KNIbP 9. AI Engineer Handbook https://t.co/vgssSOXBOA 10. AI Engineering Field Guide https://t.co/awORlkdpnZ
- 24
@ChrisGPT ·
Greg Brockman on OpenAI’s formal definition for AGI: It goes beyond a strict set of rules, noting that "everyone has their own intuitions about what AGI is." He personally says: , "I think we're about 80% of the way there." He highlights the remarkable abilities of current models like GPT 5.5 particularly in coding, admitting that "they're certainly more capable than I am at writing software." Then goes onto ask the audience , "Does anyone here feel better at writing software” than the current GPT model ?"
Watch video
- 25
@VaibhavSisinty ·
I've been tracking AI pricing models closely. Not as an observer. We spend 50 lakhs a month on AI tools as a company. When you're spending that much, you stop looking at comparison charts and start understanding how pricing actually works. And almost everyone is debating the wrong number. Anthropic's Fable 5 costs $10 input and $50 output per million tokens. GPT 5.6 Sol costs $5 input and $30 output. Kimi K3 charges roughly $3 and $15. Chamath simplified it on All-In: one million tokens as one barrel of intelligence. But published API pricing is not what anyone actually pays. Think of it like MRP. During Diwali, crackers have a Maximum Retail Price printed on the box. Some ridiculous number. The price you actually pay is a fraction of that. API pricing is the MRP of AI. The number on the box, not the number on the bill. What's actually happening: → OpenAI's subscription is becoming a token bundle. Close to $100 a month with API access baked in. Per-token math disappears. They want you price insensitive. → Anthropic and OpenAI are throwing free credits at startups. The credits are the hook. The lock-in is the business model. → Enterprises negotiate rates that look nothing like the rate card. The published price is for people who don't know to ask for a deal. Real example from our production: one run of our content plugin on Fable costs $3-4 on the API. On a subscription, that cost is absorbed. Same model. Same output. Completely different economics. Now here's where open source gets interesting and where the pricing debate misses two things: → Speed. Kimi K3 takes 25-35 minutes on tasks GPT 5.6 Sol finishes in under 10. In production, speed is cost. A developer waiting 35 minutes is burning time that never shows up on the API bill. → Compute isn't free because the model is free. Kimi K3 ran out of GPUs 48 hours after launch. Open source gives you the weights. Not the servers. The plugin that costs $4 on Fable's API runs for pennies on an open-source model when it doesn't need frontier reasoning. That routing saves us more than most companies spend on AI in a year. Stop comparing API charts. Start comparing total cost of getting the job done.
- 26
@rohanpaul_ai ·
OpenAI’s new target is an autonomous AI researcher that can break big problems into parts, run many agents in parallel, test ideas, read results, fix mistakes, and keep going for days with little help. ~ MIT Technology Review The bet is that coding agents are the first real proof this can work, because software engineering already looks like research in miniature: long chains of decisions, failed attempts, tool use, memory, and step-by-step checking. So the jump from Codex to a research system is not about one smarter answer, but about building a machine that can plan, delegate, verify, and recover across a full project instead of a single prompt. --- technologyreview .com/2026/03/20/1134438/openai-is-throwing-everything-into-building-a-fully-automated-researcher/
- 27
- 28
- 29
@sentient_agency ·
this is the most underrated AI resource on the internet. OpenAI quietly maintains a free repository of working code recipes for every single thing you can build with their models, and almost nobody talks about it. It's called the OpenAI Cookbook. Hundreds of recipes. Updated weekly. Open source. Here's what's actually inside: → Full RAG pipelines with vector databases wired end to end → Function calling and tool use patterns for real agents → Fine-tuning workflows with dataset prep and evaluation → Vision recipes for OCR, document parsing, and image analysis → Realtime API patterns for voice agents and live streaming → Embedding strategies for semantic search at scale → Structured output recipes that never break your JSON schema → Batch processing for cutting API costs by 50% → Audio recipes for transcription, translation, and TTS Every recipe is a runnable Jupyter notebook. Every line of code works. Every pattern is production-tested. https://t.co/A0x1rbnVMb https://t.co/OCbaYm17WX The same engineers building OpenAI's API write these recipes. You're getting the internal patterns for free.
- 30
@milesdeutscher ·
GPT-5.6 first impressions (raw thought dump): • Better than Fable? From my tests so far, Fable still seems like the king of raw intelligence and true autonomous work. However, GPT-5.6 models are hands down the better "daily drivers" due to their pricing/speed. I'll be using a combo of both to maximize intelligence and cost efficiency. • Thoughts on GPT Work This is obviously OpenAI's attempt to create a competitor to Cowork. No doubt GPT Work is powerful, but for me, I like how I have Cowork set up, and I'll just stick with it for now, since all my workflows are there and I like the other features (Dispatch, Plugins, my Claude Skills, etc.). • My favorite model from the bunch GPT-5.6 Sol. It's insanely autonomous and seems to just push through to get work done - no matter the blockers. • Frontend improvements Massive improvement across the board in frontend/design work - it seems like over the past month, most AI labs have made great improvements here. I think the typical AI slop frontend days are behind us. • Thoughts on hosted sites My team and I will be using the new hosted sites to share ideas. Definitely a cool way to quickly build and share ideas, and I think if you're a small team/department, you'll really like this feature. Overall - only a few hours since release, but this is generally what I expected. Going to plan some content for you guys now. If you have any requests or want to see anything in particular, leave it in the comments below - I'll read everything. ChatGPT Work Ultimate Guide? GPT-5.6 trading bots? Let me know!
- 31
@eric_seufert ·
An alternate information space seems to have developed around the chatbot advertising market that has led people to believe that OpenAI's advertising product is an abject failure, ads in chatbots are fundamentally nonviable anyway, and the entire category will never generate a meaningful amount of revenue. Meanwhile, in the primary information space, we know that Alphabet saw Search queries reach an all-time high in Q1 while Search revenue grew by 19% and ads in AI Overviews monetize at parity with those in traditional Search. The company is expanding advertising in its AI-generated and chatbot surfaces cautiously but deliberately: it introduced formats like Conversational Discovery ads and Highlighted Answers in AI Mode in May of this year. The latest usage numbers shared by Google are 2.5BN MAU for AI Overviews and 1BN MAU for AI Mode. AI Overviews is roughly two years old; AI Mode is roughly one year old. Further, we know that OpenAI reached $100MM in annualized advertising revenue a mere six weeks after introducing advertising with what was essentially a manually operated product. Since then, OpenAI has executed at breakneck speed to introduce new features: CAPI and pixel integrations in May and Custom Audiences last month. ChatGPT now serves more than 1BN MAU and was already serving more than 900MM WAU in March. ChatGPT launched at the end of 2022. OpenAI's advertising revenue goal for 2030 is incredibly ambitious; I have no affiliation with OpenAI, and I don't have any quantitative indication of whether they'll meet that goal. But to conclude that OpenAI's advertising efforts are somehow failing now, or that the market won't evolve into something meaningful, feels to me totally detached from the primary information space.
- 32
- 33
@rohanpaul_ai ·
📢 OpenAI will nearly double its workforce as it pivots away from consumer experiments toward a massive push into the business market. The company wants to stop competitors from taking over the corporate space by putting thousands of new engineers and sales specialists directly into the field. This hiring spree aims to bring the total headcount to 8,000 employees by December-26. Leadership recently issued a code red to pause non-essential projects and focus entirely on making ChatGPT a better tool for professional work. Most of the new hires will focus on product development and a strategy called technical ambassadorship. These technical ambassadors work like on-site experts who help big companies install and customize AI models for their specific business needs. OpenAI is also planning to merge its coding model, Codex, with ChatGPT into one single app to better serve desktop users in the office. --- reuters .com/business/openai-nearly-double-workforce-8000-by-end-2026-ft-reports-2026-03-21
- 34
@varunram ·
I’ve talked to atleast 5 founders in the past two weeks who shifted six figure spends from Anthropic to OpenAI because they distrust Anthropic. If OpenAI becomes a trusted model source and companies don’t have to worry about them encroaching on app layer, it’s a no brained Also OpenAI has been clear (now and historically) in what they want to achieve. Anthropic is very unclear on this app layer vs model layer distinction. OpenAI was the first to open up GPT Store (great idea but a little too early)
- 35
@ai_for_success ·
🚨 OpenAI has released GPT-Rosalind, a frontier reasoning model built to accelerate life sciences research and drug discovery. Available as a research preview via ChatGPT, Codex, and API This looks big from OpenAI.
Watch video
- 36
@ishuagra02 ·
The Anthropic + OpenAI rivalry is so interesting because one’s mistakes leads to the other’s success. Two months ago, Anthropic was surging in App Store charts, hot off OpenAI’s deal with the Department of War. But the past month, they’ve slowly eroded the trust that they built up; > users saw faster token consumption without warning > frequent Claude downtime (only 2 nines of availability) > reports of Opus 4.6 intelligence dropping > Opus 4.7’s adaptive thinking can’t be disabled > Claude Code became restricted to Max subscribers due to a misconfigured “test” > Mythos got accessed by unauthorized users All of this has led to Codex crossing 4 million users, with one million in the last 2 weeks. And with their new image model and GPT 5.5 on the horizon, OpenAI has been shipping great updates that has won trust back. I’m not sure what’s going on at Anthropic but if Mythos isn’t as good as the benchmarks suggest upon release, they’ll be far behind.
- 37
- 38
@DanielMiessler ·
So one thing I find really interesting right now is the difference in how it seems OpenAI and Anthropic are approaching the market. I see Anthropic as primarily, this is an oversimplification, but I think Anthropic is primarily going after the enterprise and trying to be all the internal workflows, all the unified entity context. Kind of almost like a Palantir-type situation of understanding the business and helping the business actually perform operationally. I feel like that is the massive force of what they're pushing into. Obviously, they care about consumer as well, so both are doing both. But I feel like OpenAI is largely focusing on the bigger consumer vision. So, I opened ChatGPT the other day, and I noticed it said, "Hey, connect your Apple Health." So I feel like OpenAI is going after all the personal stuff, and they hired Jony Ive just as another piece of sort of evidence here. They hired Jony Ive to build a piece of hardware, which I think, and kind of the interpretation is basically that's meant to replace the iPhone or the phone in general. Like, to have an AI device where you're talking to your AI. And Sam has been talking about this massively. Dario seems to talk more about the future and work and the enterprise. Again, with some exceptions. And Sam seems to talk more about the future of like the personal ecosystem, having an agent always available. And obviously, given what I talk about and my sort of orientation, I really find that OpenAI thing quite compelling. Obviously, I like both. I'm really interested in both. But I feel like OpenAI is more aligned in this particular case with the way I've seen AI going, which is you have your personal assistant, that is who you're talking to. The personal assistant is then doing all the different things for you in all these different places, and basically operating on your behalf. And with the merging of ChatGPT and Codex into the single app, and the more and more connectors that OpenAI is bringing in, it really does feel like they are definitely heading in this consumer sort of orientation. And what I find interesting about this is both are multi-trillion dollar businesses, right? Replacing the operations layer of all these businesses is just extremely lucrative, right? And that's where Anthropic is sort of heading. But also just becoming the single agent, becoming Her or becoming Jarvis for all of humans, right, is kind of the TAM for OpenAI here, in this particular lens. And that is also a multi-trillion dollar market, right? So I feel like both are doing something extraordinary, really huge, that the other is kind of not touching as much, and I just find that really interesting. Curious to hear if you see it differently, or if you kind of agree with this approach.
Watch video
- 39
@WesRoth ·
OpenAI’s revenue reportedly accelerated again in July. CFO Sarah Friar told employees that the company’s annualized recurring revenue exceeded the level recorded during the entire second quarter. The exact number was not disclosed. But OpenAI says three products are driving the momentum. GPT-5.6, ChatGPT Work and Codex. Board chair Bret Taylor also admitted that OpenAI entered the year playing catch-up to Anthropic in AI coding. Now, he says, Codex is gaining traction. And some Claude Code users are reportedly looking for alternatives after receiving unexpectedly high bills. That may become one of the most important battles in AI. Not just which model writes better code but which one delivers enough value to justify the cost of using it every day. OpenAI already has massive consumer distribution through ChatGPT. Now it is turning that audience into a platform for coding, agents and professional work.
- 40
@theinformation ·
OpenAI’s Codex has become one of the company’s fastest-growing products, with enterprise revenue recently rising 50% week over week. Now OpenAI is folding it into ChatGPT as it races Anthropic to turn coding agents into general-purpose work tools. Full story: https://t.co/91x9gjWSWx
- 41
@CodeByNZ ·
“Elon Musk: I am the reason OpenAI exists. It wouldn’t exist without me.” What makes that statement so interesting is how different OpenAI looks from the vision he originally described. The idea was supposed to be: > open-source > nonprofit > and the opposite of Google’s closed AI dominance. That’s literally where the name “OpenAI” came from. Now the company is raising tens of billions, building closed frontier models, partnering heavily with Microsoft, and becoming one of the most powerful centralized AI companies on earth. The AI race moved so fast that even the labs created to prevent concentrated AI power slowly became concentrated AI power themselves.
Watch video
- 42
@HarryStebbings ·
The AI talent wars have never been crazier. Within 48 hours, Google lost two of their generational scientists to OpenAI and Anthropic. I sat down with @jasonlk and @rodriscoll to discuss it, along with the biggest news in tech this week: - Deepseek Raises $50BN - Wall St's $725BN AI Question - The Rise of Open Source & How it Threatens OpenAI & Anthropic - OpenAI Builds its Own Chip: Jalapeno My notes below: 1. The Number Three Closed-Source LLM Is Most at Risk In the closed-source foundation model race, the number three vendor faces severe pressure as multi-model routing spreads across enterprises. Historically, a third-place software or cloud vendor could survive by being cheaper or simpler. But today, that tier is being squeezed by highly capable open-source models, many heavily subsidized by China, making developers less likely to care about a closed-source number three. 2. The Playbook for Building a Startup to Its First 100 Employees Has Changed Scaling a startup to its first 100 or 200 employees has changed radically from the remote-work era. In today’s hyper-competitive AI landscape, companies built around relaxed schedules and 20-hour remote workweeks will not win. The modern model is lean, elite, highly compensated teams working intensely in person, often six or seven days a week, to survive nonstop product sprints. 3. Why It Makes No Sense for OpenAI and Anthropic to Build Their Own Chips Frontier AI labs have captured the greatest demand engine in modern technology and should focus entirely on winning enterprise customers. Designing custom hardware is a massive distraction when cloud giants like Oracle, Google, Microsoft, and Amazon are already competing aggressively to provide cheap compute. Those infrastructure vendors absorb the capital risk, letting labs prioritize product growth and customer adoption. 4. Why Vertical Integration Makes No Sense for OpenAI and Anthropic Closed-source labs scaled rapidly by staying asset-light and outsourcing heavy infrastructure needs. Their model worked because cloud hyperscalers absorbed hundreds of billions of dollars in cumulative CapEx on their behalf. Forcing backward vertical integration down to the chip level would introduce massive capital liabilities and undermine the operational efficiency that made the model work. 5. Show Me the ROI Next Year Enterprise software is moving from loose AI experimentation to strict financial accountability. Early corporate budgets funded unconstrained “token maxing” to build basic AI fluency, but the 2027 narrative will demand measurable ROI. CIOs will no longer allocate tokens based on compelling pitches alone; they will require verified departmental efficiency gains or clear revenue growth. (links below)
- 43
@geoffwoo ·
how to think about OpenAI going IPO mode: founders should stop treating the frontier lab like a weather report and start treating it like a gravity well. when GPT-5.5, GPT-6, spud, mythos keep pulling capability upward, thin product surfaces get weird. APIs survive if they own distribution, proprietary workflow context, regulated trust, or a budget line that moves. everything else becomes a nice demo orbiting the sun.
- 44
@Suryanshti777 ·
OpenAI's own coding agent just got forked into a weapon against OpenAI's own business model. Open Interpreter took Codex, ripped out the parts that only work well with GPT, and rebuilt it to squeeze max performance out of open models instead — Kimi K3, DeepSeek, Qwen, GLM. Type `/harness` and switch the entire agent behavior mid-session. Same interface, different brain underneath. It also speaks the actual Codex exec protocol. One line swap in your existing SDK code and you're running open models through the same pipeline you built for Codex. 66.8k stars. Apache-2.0. Fork it, ship it, no permission needed. The lesson here isn't "open models are catching up." It's that the harness — the thing wrapped around the model — might matter more than the model itself. And that part just got forked wide open. Link in comments 👇
- 45
@jalaal_tweets ·
OpenAI just released a model you probably can't access. That's the point. GPT-5.4-Cyber dropped yesterday. It’s gated to vetted security researchers only through their Trusted Access for Cyber programme. No public API. No waitlist you can join. Anthropic did the same thing last week with Claude Mythos Preview. Two frontier labs are now deliberately building AI that most builders will never touch. Here's why that matters if you're in Web3: The security infrastructure underneath DeFi (the systems that audit smart contracts, detect exploits before they happen, flag suspicious on-chain behaviour) is quietly being rebuilt by AI models that exist behind identity verification and institutional vetting. You won't use these tools directly. But the protocols you use will. The auditors you trust will. The wallets protecting your funds will. This is a new category of AI development. Not democratised, not open-source, not accessible. Deliberately locked. And it's moving fast. Most AI commentary focuses on what you can access. The more interesting story is what you can't and who's building with it right now. The gap between institutional AI and builder AI just got wider.
- 46
@shawnchauhan1 ·
OpenAI just acquired the team behind Python's most popular developer tools. uv, Ruff, ty - the tools serious Python engineers actually use. Codex already crossed 2 million weekly active users. Usage up 5x since January. This acquisition is not about writing code. It is about owning the environment around the code. Package management. Quality checks. Project maintenance. The full workflow. OpenAI is not building a smarter autocomplete. It is building the operating layer for how software gets made. That is a different and much larger category.
- 47
@_vmlops ·
OPENAI HAS A SECRET WEAPON AND IT'S FREE the openai-cookbook repo is lowkey one of the most underrated resources on github ▫️ real code examples using the openai api ▫️ covers gpt-4, chatgpt, embeddings, agents, and more ▫️ written in python (but concepts apply anywhere) ▫️ actively maintained with 1,367+ commits no more guessing how to structure api calls everything from rag pipelines to function calling it's already in there with working notebooks if you're building with ai and not using this, you're making things harder than they need to be https://t.co/m1iy7IO6ls
- 48
@PlanX_Lex ·
Hot take on the OpenClaw story: When OpenAI hired OpenClaw’s founder, Peter Steinberger, the project itself was open-sourced shortly after. That sequence is a signal. In tech acquisitions, companies rarely open-source something that represents a real strategic moat. They open-source things that are: • useful • interesting • but not defensible OpenClaw is essentially an agent orchestration layer. It connects: LLMs + tools + messaging interfaces + local execution. Useful? Yes. Hard to replicate? Not really. The core primitives behind it already exist everywhere: model APIs tool routing context windows workflow graphs That’s why the real asset in this story was likely the talent, not the software. OpenAI hired the founder. The code became open source. This is a pattern we’ve seen many times in infrastructure software: if something is strategically critical, it stays proprietary. If it’s easy to rebuild, it becomes ecosystem infrastructure. Which leads to a bigger point about AI systems. The long-term moat will not come from thin orchestration layers. It will come from: domain intelligence evaluation systems proprietary data loops vertical models and risk-aware execution frameworks. General-purpose AI agents may generate hype. But real defensibility in AI infrastructure will emerge where decision intelligence and domain-specific learning live. That’s where the real competition will be. @openclaw @OpenAI #AI
- 49
@boyuan_chen ·
Distribution. OpenAI just shipped a Codex plugin that runs inside Claude Code. You can trigger Codex for code review, adversarial review, or task delegation without leaving Anthropic's CLI. Piggybacks on your existing ChatGPT subscription. Think about what this means. OpenAI gets usage data, billing, and developer habit from inside a competitor's tool. Every Codex call from Claude Code feeds OpenAI's flywheel while the developer thinks they're using Anthropic. Model benchmarks decide who's best this quarter. Workflow integration decides who's default next year. https://t.co/X6LfmGg6Ss
- 50
@ashmaurya ·
A founder I coached built a productivity tool on top of GPT-4. Clever prompt layer. 8 months in. One Friday, OpenAI shipped a feature that did the same thing for free. The assumption that "API access = unfair advantage" was decision debt. Nobody had tracked it. Nobody had contradicted it.
Best Tweets by Topic