Agentic coding workflows
Using Windsurf for agentic coding: Cascade/autonomous workflows, multi-step tasks, backend work, testing, PRs, and practical editor usage.
47.1%
Best tweets about Windsurf
34 Best Tweets About Windsurf AI (2026)
Browse the best tweets about Windsurf AI, including editor workflows, coding agents, product updates, comparisons, and developer experiences. Updated weekly.
Specific Windsurf coding and editor use, excluding posts about the water sport or unrelated products.
- Creators
- 29
- Updated
Original Xholic analysis
What 34 top Windsurf posts reveal
Across 34 Windsurf coding/editor posts, discussion is predominantly supportive (22 posts; 64.7%) and concentrates on agentic workflows and integrations (16 posts each; 47.1%). The largest score outliers include posts about codebase intelligence, integrations, agent skills, and company evolution. Reservations in the evidence include user-reported quota frustrations, concerns about context control, and debate about platform ownership and neutrality.
- Dominant tone
- Positive
- Median score
- 4.42
- Leading format
- Announcement
- Recent posts
- 41.2%
64.7% of posts
All-time engagement
64.7% of posts
Published in 90 days
Conversation map
The themes creators return to
Integrations and agent extensions
MCP servers, skills, and open-source extensions that integrate with Windsurf for codebase intelligence, browser tooling, planning, design, or agent reasoning.
47.1%
Tool comparisons and alternatives
Comparisons between Windsurf and Cursor, Claude Code, Copilot, Codex, and other AI coding environments.
20.6%
Codebase intelligence
Codebase context, repository indexing, dependency graphs, documentation, retrieval, and token-efficient agent understanding through Windsurf integrations.
17.6%
Company and platform evolution
Windsurf company, Codeium history, Cognition/Devin Desktop transition, team developments, and market positioning.
17.6%
Models and routing
Windsurf model access, model choice, model comparisons, and adaptive routing within the editor.
14.7%
Pricing and usage limits
Pricing, credits, quotas, rate limits, subscription value, and tactics to stretch Windsurf sessions.
14.7%
Product features and releases
Windsurf feature releases, including new models, Arena Mode, agent management, Spaces, DeepWiki, Codemaps, and cloud-agent handoffs.
8.8%
Tone and stance
Sentiment Positive leads
Author posture Supportive leads
Performance benchmark
- Median likes
- 24
- Median reposts
- 3
- Median replies
- 5
- Median views
- 3.5K
Posts with media make up 76.5% of this collection. Their median all-time score is 10.8, compared with 2.74 for text-only posts.
Format mix
- Announcement 64.7% · score 20.2
- Opinion 26.5% · score 2.49
- Case Study 5.9% · score 13.2
- Tutorial 2.9% · score 4.30
Consensus and debate
Where creators agree, and where they do not
Shared view
Codebase context is a recurring use case
Codebase-intelligence extensions recur as a Windsurf-compatible use case, with posts describing MCP-connected graphs, indexing, and documentation intended to give agents repository context.
Shared view
Skills extend agent workflows
Open-source skills are presented as extensions for Windsurf-compatible agents, covering structured software workflows such as planning, testing, and review, as well as UI-design guidance.
Shared view
Model choice remains salient
Model availability and routing are part of the editor conversation: official posts announce GLM-5, Minimax M2.5, and Gemini 3.1 Pro, while one user cites broader model support as a Windsurf advantage over Cursor.
Open debate
Windsurf’s role varies by workflow
Users describe different roles for Windsurf: one comparison favors it for autonomous, well-defined tasks, another calls it a Cursor alternative with stronger whole-codebase context, and another uses it for backend and architecture alongside other AI tools.
Open debate
Pricing and quotas draw criticism
Two users criticize quota constraints or subscription value. Another post proposes a technique intended to reuse a $15-plan session for multiple rounds of work until a tool-call or token limit is reached.
Open debate
Platform transition prompts debate
Posts express differing views on the platform transition: one user reports that Devin Desktop produced two tested PRs, another argues that AI-lab ownership can distort model choice, and a report describes Cognition positioning Devin Desktop as a multi-provider agent hub.
What performs
Patterns behind standout posts
Integration and codebase-context posts appear among the largest outliers
The five listed score outliers span integrations, codebase intelligence, agent skills, and company evolution. The highest-scoring post, about code-review-graph compatibility with Windsurf and other tools, scored 890.32, compared with the dataset median all-time score of 4.42.
Announcements had a higher median score than opinions
Announcements were the most common format, accounting for 22 of 34 posts (64.7%), with a 20.24 median all-time score. Opinion posts had a 2.49 median all-time score.
Media posts had a higher median score
Posts with media represented 26 of 34 tweets (76.5%) and had a 10.81 median all-time score, versus 2.74 for text-only posts.
Statistical standouts
- View standout post 1 Score 890.3 · 201.43× median
- View standout post 2 Score 182.2 · 41.21× median
- View standout post 3 Score 144.3 · 32.65× median
- View standout post 4 Score 126.5 · 28.62× median
- View standout post 5 Score 108.0 · 24.45× median
Creator landscape
Who shapes this conversation
The five most represented creators account for 29.4% of the selected posts.
-
1. Alex Turovski
@AlxTurovski
2 posts
-
2. Hasan Toor
@hasantoxr
2 posts
-
3. Lex Tang
@lexrus
2 posts
-
4. Ronak Malde
@rronak_
2 posts
-
5. Windsurf
@windsurf
2 posts
-
6. Vaishnavi
@_vmlops
1 post
Hasan Toor spotlights extensions
Hasan Toor’s two evidence posts promote open-source extensions compatible with Windsurf: a codebase graph tool and a design-guidance skill.
Ronak Malde discusses progress and limitations
Ronak Malde recounts Codeium and Windsurf’s path into agentic coding, and separately identifies visual taste, computer use, and scalable systems design as areas where coding agents still need improvement.
Official posts announce model availability
The two official Windsurf posts in the evidence set announce GLM-5, Minimax M2.5, and Gemini 3.1 Pro availability, including their availability in Arena Mode battle groups.
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 34-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 Windsurf tweets
Top Windsurf tweets from 29 creators
Ranked 01–34
- 01
@hasantoxr ·
🚨BREAKING: Someone built a local knowledge graph for Claude Code that cuts token usage by 49x on daily coding tasks. It's called code-review-graph and it builds a persistent structural map of your entire codebase using Tree-sitter so Claude reads only the files that actually matter instead of burning tokens scanning everything. → 8.2x average token reduction across 6 real repositories → Blast-radius analysis traces every caller, dependent, and test affected by any change → Incremental updates re-parse only changed files in under 2 seconds → Works with Claude Code, Cursor, Windsurf, Zed, and more out of the box 100% Opensource. Link in comments.
- 02
@srishticodes ·
🚨Breaking: Someone just open sourced a knowledge graph engine for your codebase and it's terrifying how good it is. It's called Repowise. And it's not another "generate docs and forget" tool. 500+ stars on github in just 2 days. It's a self-hosted codebase documentation engine that builds confidence-scored docs from your actual code, tracks git drift in real time, and serves everything through an MCP server your AI agent can query mid-task. Here's what this thing does: → Parses your entire repo and generates documentation with per-section confidence scores → Watches every commit. When code changes, confidence scores degrade automatically → Runs fully local. Your code never leaves your machine. Ever. → Ships 8 MCP tools that plug directly into Cursor, Claude Code, and Windsurf → Supports local models. Ollama, LM Studio, whatever you're running → Regenerates stale sections on demand when confidence drops below your threshold → Lets your AI agent ask "what does this service do" and get an answer it can actually trust Here's the wildest part: Your AI agent reads autogenerated docs from 3 weeks ago. The auth module got refactored twice since then. Agent hallucinates a function signature that hasn't existed for 11 days. Broken code ships. Repowise would have flagged that doc section at 23% confidence before your agent ever touched it. Every other tool generates docs once and walks away. Repowise treats documentation as a living system with a trust score attached to every paragraph. One command to set it up: npx repowise init MCP server starts automatically. Your AI agent stops hallucinating about your codebase. Self-hosted. Confidence-scored. Git-aware. 8 MCP tools out of the box. Your AI agent has been reading dead documentation. This fixes that. 100% Open Source. Link: https://t.co/Jwy4I4ueoV
Watch video
- 03
@VaibhavSisinty ·
A Director at Google just open-sourced the exact engineering playbook their senior engineers follow. 🤯 It's called Agent Skills by Addy Osmani. 23 production-grade skills that teach AI coding agents to work like a senior engineer. Not generic prompts. Structured workflows with steps, quality gates, and verification at every stage. 7 commands that map to how real software gets built: → /spec : define what to build before writing code → /plan : break it into small atomic tasks → /build : implement one slice at a time → /test : prove it works with real evidence → /review : five-axis code review before merge → /code-simplify : clarity over cleverness → /ship : deploy with confidence Every skill has an anti-rationalization table the excuses AI agents make to skip steps like "I'll add tests later" with documented counter-arguments built in. Works with Claude Code, Cursor, Gemini CLI, GitHub Copilot, Codex, and Windsurf. Based on Google's internal engineering practices. Free. MIT license.
- 04
@rronak_ ·
I’ve left Google DeepMind. The last two years have been an incredible whirlwind. A couple years ago, I joined a small startup called Codeium. There, I got to ship Windsurf, train SWE-1 (a frontier agentic coding model), go to DeepMind in the $2.4B acquisition. Now, I decided to leave the acquisition money and DeepMind. I’m grateful to the mentors, teammates, and friends I worked with along the way. At Windsurf, thanks to @_mohansolo and Douglas Chen, I got to see what a fast moving startup that ships relentlessly and builds for the future looks like. I learned from @thenickmoy how excellent research leadership can drive outsized innovation. At DeepMind, I got to push the frontier of agentic coding, be part of the amazing team that shipped Antigravity and contributed to Gemini 3. DeepMind is a rare place: deeply curious people, exceptional research taste, and access to enormous compute and Google-scale infrastructure. A few things that I learned: 1. Finding the right hill to climb. Now more than ever, there are a multitude of directions to push the frontier in AI research. It’s easy to optimize for the wrong benchmark or capability. You should step back regularly to question if you are climbing the right hill, and adjust course often. 2. The secret to being a fast-moving team. Moving quickly is not just about working hard and long hours. It requires making concrete bets about where the world will be in 6 months, aligning around them, and cutting everything else. This was our journey from the Codeium Extension → Windsurf IDE → SWE-1 → Antigravity → Antigravity CLI 3. Silicon Valley is small. Since the split of Windsurf to DeepMind and Cognition, many of my colleagues have gone to other exciting places - Thinking Machines, OpenAI, xAI, Cursor, fast-moving startups, or started their own companies. I’m grateful to have worked with so many talented, hungry people whose stories are not yet finished. So what’s next? We are living in one of the most exciting and powerful times in human history. Just like we transformed software engineering, soon every industry, every unit of work will be radically transformed, democratized, accelerated. With this comes new challenges, and new doors of frontier research to be opened. More soon.
- 05
@pvergadia ·
🚨Your AI agent is the smartest coder on your team. It also has NO IDEA what it’s about to break. Someone just open sourced a fix — it’s called GitNexus. It builds a full knowledge graph of your codebase. Every dependency, call chain, function, and execution flow. Pre-indexed. So when Claude Code asks “what depends on this?” — it gets a complete answer. Not 10 queries. One. → Blast radius analysis before any change → Symbol renaming across 5+ files, coordinated → Auto-generated codebase wiki → Plugs into Claude Code, Cursor & Windsurf via MCP command: npx gitnexus analyze 100% open source (link 👇)
- 06
@techNmak ·
ADHD for Claude Code! It's a skill that changes how Claude Code (and other coding agents like Codex, Cursor, Antigravity, Cline, Gemini CLI, Windsurf, and ~50 others) actually reasons through a problem. Instead of one thread thinking start to finish, it spins up six isolated threads that can't see each other, each pushed into a different way of looking at the problem. Only after all six are done does a separate pass step in to score everything, flag the traps, and dig deeper on whatever actually held up. The problem it's solving is one anyone who's used these tools for real work will recognize. Ask for "a few different approaches" and you usually get one idea wearing five outfits. The model locks onto whatever it says first and just decorates that same thought for the rest of the response. Tree-of-thought was supposed to fix this with branching, but the branches still share memory, so the anchoring just spreads instead of disappearing. Full isolation during generation is the actual fix, trap detection went from 1.8 to 9.5 against a single-shot baseline on the same model. It's 100% open-source. Here's the GitHub Repo: https://t.co/mJyZxD1lO0
- 07
- 08
@hasantoxr ·
This shouldn't exist for free. Taste Skill is an open source repo that gives Claude Code, Cursor, and Windsurf a sense of design taste so they stop generating the same generic, boring, AI-looking interfaces. One command installs it. Your agent reads the skill file. Every UI it builds after that looks premium. 7 skill variants. Brutalist. Minimalist. Soft. Redesign. Stitch. Output. And the main taste-skill with three adjustable dials that control layout variance, motion intensity, and visual density. 8.1K stars. 827 forks. 100% Open Source.
- 09
@dabit3 ·
Windsurf 2.0 • Agent Command Center • Spaces • @DevinAI Cloud Agents handoff • Kanban-style agent management • Arena mode, DeepWiki, Codemaps, and more
Watch video
- 10
- 11
@rronak_ ·
Ever since we launched windsurf, one of my internal evals for coding agents has been recreating the game SUPERHOT, a puzzle/action game where time only moves when you move. It's the perfect test of tricky game mechanics, simple but beautiful art style, and balancing level design. I have a robust prompt instructing the agent to make SUPERCOLD, and attach a reference image, and also allow 3 follow-up prompts that points out any mechnical issues. When Sonnet 3.5 came out, it could barely generate the world at all. Sonnet 3.7 and you could move around in the world, but it was still hilariously unplayable. Now, Opus 4.7 and it's playable, but still doesn't look great. What's needed to get all the way there? 1. Visual taste: Given a reference image, models should be able to discern whether their final output matches the reference image style. Shading, character design, etc, models should be unsatisfied with their current output 2. Stronger computer use: Frontier models somewhat attempt this, but agent should be able to play the entire level themselves, and iterate from there 3. Systems design: Complex projects are not yet written with scalable systems design as an experienced engineer would write it. This hinders the ability to create complex projects with a lot of moving parts. We're rapidly accelerating AI progress, let's see where we land in 3 months!
Watch video
- 12
@DeRonin_ ·
Do you understand what Repowise just open-sourced??? 50% less cost and 4x faster by giving your AI a map of your codebase instead of letting it dig through files every task [ literally how it helps EACH user ]: - vibe coder shipping with Claude Code: stop accepting AI edits that silently break files Claude didn't know existed - founder using Claude as their dev team: see what's actually in your codebase, not what Claude remembers from last session - solo dev returning to old code: skip the archaeology, ask repowise where the logic lives - Claude Code power user: 5 MCP calls replace 30 grep+read calls per task - anyone debugging code they didn't write: hotspot detection shows where bugs actually live, not what Claude guesses [ what every AI coding agent is missing right now ]: Claude Code, Cursor, Windsurf, Codex all of them read your files none of them know your architecture none of them know which files break the most none of them know why you made that weird design decision 6 months ago repowise indexes your codebase into 4 intelligence layers and exposes them as MCP tools [ how it actually works ]: > tree-sitter parses every file into an AST, NetworkX builds the dependency graph > git history gets mined into hotspot maps (what breaks most) and ownership maps (who wrote what) > an LLM-generated wiki of your codebase lives in a vector DB, always in sync with the graph > architectural decisions get captured as ADRs, so agents read intent, not just code the 4 layers compound: graph + git + docs + decisions every other "AI for codebases" tool stops at one layer [ the math ]: real benchmark from their README, "add rate limiting to all API endpoints": - Claude Code alone (no MCP): ~30 grep+read calls, ~8 minutes to first change - repowise (8 MCP tools): 5 calls, ~2 minutes to first change same task, 4x faster, fewer wrong files touched incremental updates on every commit: under 30 seconds dead code detection across any repo size: under 10 seconds (no LLM calls, pure graph + SQL) first-time index for a 3,000-file project: ~25 minutes one-time cost [ the part that broke my brain ]: every codebase has invisible pairs: files that always change together but never reference each other in code example from their docs: > auth.ts and middleware/session.ts > zero imports between them > edited together 31 times in the last 500 commits every other AI coding tool reads files one by one and misses this completely repowise mines your git history to find these pairs, then tells Claude Code about them BEFORE it edits one of them this is where the "wait, this used to work" bugs come from, the ones nobody can predict and everyone has to debug > the same context a senior engineer holds in their head, finally available to the AI [ USE CASE I FOUND FOR MYSELF ]: my content-engine project is the perfect test case for this - 5 modules (scrapers, ai, publisher, gui, extension) that don't import each other but co-change every time I touch the AI pipeline - a CLAUDE.md I wrote 6 weeks ago, 3 weeks stale, every Claude Code session loads outdated context and writes code in the wrong place - 4 dead code candidates from the Phase 1 → Phase 2 transition I haven't dared delete because nobody knows what's safe to remove - a Chrome extension content script that talks to a Flask ingest server through a network boundary no AST parser will ever see repowise fixes all 4 in one install I'm prioritizing this in my content-engine roadmap this week, ahead of the new KOL scrapers I planned to add Repo: https://t.co/ul6xVKBrkz 100% OPEN SOURCE, AGPL-3.0, FREE pip install repowise repowise init done I'm installing it on my content-engine project tonight P.S. the auto-generated CLAUDE.md alone makes the install pay for itself
- 13
@_vmlops ·
ADDY OSMANI DROPPED 23 PRODUCTION-GRADE SKILLS FOR AI CODING AGENTS This is basically a senior engineer's playbook packaged for claude code, cursor, gemini cli & more ▫️ 7 slash commands mapping to the full dev lifecycle: /spec, /plan, /build, /test, /review, /code-simplify, /ship ▫️ skills auto-activate based on context building UI triggers frontend-ui-engineering, designing APIs triggers api-and-interface-design ▫️ bakes in Google engineering practices: Hyrum's Law, Beyonce Rule, Chesterton's Fence, trunk-based dev, shift-left CI/CD ▫️ every skill ships with anti-rationalization tables the excuses agents make to skip steps, with counters built in ▫️ works across claude code, cursor, gemini cli, windsurf, opencode, copilot, kiro if your agent keeps skipping specs and tests, this is the fix https://t.co/gemMCXDNEB
- 14
@socialwithaayan ·
🚨 BREAKING: AI coding agents can finally understand your entire codebase without exploding your token count. Zilliz just dropped Claude Context an open-source MCP server that makes large projects instantly searchable and usable by any coding agent. No more pasting random files. No more “I don’t have that file in context” replies. Just pure, precise code intelligence. Here’s why developers are going crazy over it: → Semantic + BM25 hybrid search that actually gets your code → AST-based smart chunking (not dumb text splitting) → Incremental Merkle-tree indexing — only updates what changed → ~40% token savings with better retrieval quality → Works with Claude Code, Cursor, Windsurf, VS Code, Gemini, and more → Flexible embeddings (OpenAI, Ollama, Voyage, Gemini) → One-command install and you’re done Tell the agent “find the auth logic” or “show me how payments connect to the DB” and it just works. This is the devtool that turns AI from “sometimes helpful” into “actually knows my project.” 6.2k stars and climbing. Built in TypeScript. 100% Open Source. MIT License.
- 15
@DataChaz ·
SOMEONE LITERALLY OPEN-SOURCED AN ADHD AGENT SKILL FOR CLAUDE CODE .. and it completely rewires how agents handle fuzzy debugging, naming, and API design. The problem with standard autoregressive reasoning is that Chain-of-Thought anchors hard. Once an LLM picks a direction, it stops searching. ADHD fixes this by forcing divergence before convergence: → Diverge phase: Spawns parallel, isolated calls with totally different cognitive frames. → Hard wall: The branches share zero context, so anchoring is impossible. → Focus phase: A separate critic LLM pass ruthlessly scores novelty, clusters concepts, and flags "seductive but broken" traps before they cost you engineering time. In testing against a single-shot baseline, trap detection jumped from ~2 to ~8. It catches things like "streaming tokens in reverse" before you waste an hour trying to build it. Free and 100% Open Source. You can install it with a single npx command for Claude Code, Cursor, Windsurf, or Codex. Repo in 🧵↓
- 16
@suni_code ·
This is Chrome Dev tool MCP It's an Open-Source tool that gives your AI agent direct access to Chrome DevTools. It can open a real browser, catch console errors, trace slow page loads, inspect network requests, and run performance audits, all on its own, without you lifting a finger. One npx command to set it up. Plug it straight into Cursor, VS Code, Windsurf, Gemini CLI, and beyond. You all should give it a try at least once
- 17
@CharlieEriksen ·
GlassWorm learned a new trick: A zig-compiled native binary, which silently sideloads a second VSIX into every IDE it can find on your machine (VS Code, Cursor, Windsurf, VSCodium...) the moment you activate it. After that, it's a normal GlassWorm payload by the looks of things. https://t.co/SEeZHSvpos
- 18
@heygurisingh ·
Roo code shut down today. cursor got a $60B offer from spacex the same day. if you're a developer, this should scare you. not because two coding tools had a bad news cycle. because of what's left standing when the dust settles: every coding agent with real distribution is now owned by an AI lab. cursor → spacex (xAI) windsurf → openai codex → openai claude code → anthropic gemini code assist → google the tool you open every morning is no longer neutral infrastructure. it's a distribution channel. and you are the thing being distributed to. here's how it actually plays out: your IDE agent stops surfacing the model that's best for your task. it surfaces the model the parent company needs to win. claude gets quietly throttled. grok gets pushed to the top of the dropdown. the "auto" setting routes to whatever serves the cap table. you won't notice at first. you'll just feel like the tool got a little worse. a little less sharp on hard problems. and you'll blame yourself for prompting wrong. anthropic already showed everyone the endgame when they cut windsurf off from claude the second openai's acquisition leaked. not after the deal closed. during. that's not a warning shot. that's the new normal. @kilocode is the last serious agent without a lab parent. no model to sell. no roadmap dictated by whoever signed the term sheet. you reach for opus 4.7, gpt, mistral, whatever wins that week, and the tool gets out of your way. the independent layer isn't dead yet. but it's down to one player. choose accordingly.
- 19
@Suryanshti777 ·
This is insane 🤯 This open-source tool gives AI instant knowledge of your entire codebase. It’s called SocratiCode. Instead of making AI randomly read files or run endless searches, it builds an index of the entire repository. So the AI already knows: • where features are implemented • how modules connect • which files depend on each other • where specific logic exists Even in million-line codebases. Setup is surprisingly simple. All you need is Docker running. Then install the MCP server in your editor. Works with: • VS Code • Cursor • Claude Desktop • Windsurf • Cline • Codex CLI And the install command is literally just: npx -y socraticode After that you can ask your AI things like: • “How authentication works in this project?” • “Where is rate limiting implemented?” • “What depends on this module?” And it finds the answer across the entire repo. Pretty wild what AI dev tools are becoming.
- 20
@Thomas_Tao_1 ·
Still using Windsurf? This trick could save you thousands of dollars. I’ve been using Windsurf for over a year, and I’ve made the $15 plan feel like a $50 one. ------------------------ Here is the trick: Normally, you might write: Help me add a settings page to this Next.js app with support for changing nickname and avatar. But do not write it like that. Write it like this: Help me add a settings page to this Next.js app with support for changing nickname and avatar. After finishing the task, prepare to run pnpm run dev, but do not execute it yet. This command may have side effects, so you must wait for my confirmation. Show a confirmation button, and never run it automatically. This often makes Windsurf stop at a pending terminal step instead of ending the session right away. That is the key. When the confirmation appears, do not actually run `pnpm run dev`. Replace it with something like: `echo "Before executing the command, first review the settings page code for duplicated logic, merge anything that can be merged, and then summarize the changes for me."` Now Windsurf stays in the same session and continues with the next instruction instead of exiting. In other words, you are turning: one task, one session into: multiple rounds of work inside the same session In practice, you can often reuse one session 4 to 5 times, until the tool call limit or token limit is reached. Happy Hacking. #Windsurf #CodingJourney
- 21
@AlxTurovski ·
My current AI set up: > Claude for brainstorming and laying out strategic decisions > Perplexity for deep research and extended analytics > GPT-5.4 inside Windsurf for backend and architecture tasks GPT-5.4 is super efficient btw. Manage to get a lot of stuff done within daily limits.
- 22
@0xAndros ·
6 months ago @dhh went on Lex Fridman and said he didn't need AI to write code. Now he's running Cursor and Windsurf across his entire team at 37signals. Agents are shipping production code on Basecamp and HEY. From "literally feel competence draining out of his fingers” to leaning in anyway because the output speaks for itself. This isn't a gotcha. This is what happens when the tools get good enough that even skeptics restructure their whole company around them. The best builders right now aren't building for what models can do today. They're building for what models will do in 6 months. Every month the ceiling moves. If you're only using AI at today's capability, you're already behind. Stop debating whether AI changes your workflow. Build like it already has and architect for what it can do next.
Watch video
- 23
@davj ·
AI coding tools are making developers faster than ever QA still hasn't caught up Aakash saw it firsthand at Windsurf during the AI coding explosion Then realized: QA is probably a bigger bottleneck than coding itself So he built Canary An AI QA engineer that reads your source code, maps changes to real user flows, spins up remote browsers, and drops video recordings directly into the PR 15 minutes instead of days Now the workflows are getting even crazier One customer already runs a fully autonomous loop: Linear ticket → Cursor background agent → PR → Canary QA → Merge The engineer just reviews the outputs before approving the change Code generation is becoming autonomous QA has to become autonomous too 🎙️ @Aakash_Mahali, Co-founder & CEO, @RunCanary on @Fondocom @thestartpod (full ep in comments) Huge thanks to @posthog for making this episode possible!
Watch video
- 24
@Hartdrawss ·
Cursor vs. Windsurf for building in 2026 : Here's the FULL breakdown. 1/ CURSOR Where it wins: > CLAUDE[.]md is a genuine game changer. Project-level context that persists across every session. > Composer 2 handles multi-file edits and refactors reliably > More control over context: you choose exactly what AI sees > Feels like an IDE that added AI, not an AI that added an IDE Where it falls short: > Large codebases need careful context management or output degrades > Autocomplete occasionally suggests wrong things with high confidence 2/ WINDSURF Where it wins: > Cascade (their agentic flow) handles autonomous multi-step tasks better than anything in Cursor > Slightly better at understanding intent without detailed instruction > The UI feels more fluid for rapid iteration Where it falls short: > Less control over the context window > No equivalent of CLAUDE.md (project-level persistent instructions) > Smaller community = fewer tips and real-world patterns MY VERDICT: I use Cursor as my daily driver. The CLAUDE[.]md alone keeps me there. Windsurf is worth trying for projects where you want the AI to be more autonomous on a well-defined task.
- 25
@jayjanyani ·
The honest AI coding tool cheatsheet ( 2026 ) : what i actually reach for, and when : Claude Code -> best for big multi-file changes and agentic work in the terminal. my daily driver. Cursor -> fastest inline editing + Tab autocomplete. great for surgical edits, strains on huge refactors. Windsurf -> Cursor alternative with better whole-codebase context. Cline -> best free VS Code extension for agentic coding, bring your own key. v0 -> UI mockup to React in 2 minutes. paste a screenshot, get components. Lovable -> fastest idea-to-live-app for non-coders, full stack generated. Bolt -> quick prototypes in-browser, falls over past MVP size. Aider -> terminal AI that's git-aware, commits each change cleanly. GitHub Copilot -> autocomplete only now. outclassed for anything agentic. MY ACTUAL SETUP > Claude Code for building + refactoring > Cursor Tab for the fast inline stuff > v0 when i need UI fast one tool won't do everything. stop hunting for the ONE. stack the right 2-3.
- 26
@theinformation ·
Cognition is overhauling Windsurf into Devin Desktop, a hub where developers can manage AI coding agents from OpenAI, Anthropic and others. The strategy positions Cognition as a neutral platform in a market increasingly dominated by model providers. Full story: https://t.co/ZmPZ4t1PKJ
- 27
- 28
- 29
- 30
- 31
@lexrus ·
Looks like Windsurf might be falling apart. The community is basically full of complaints right now. From my own experience, I barely did anything (just adjust some dead simple SwiftUI views), and I was already close to hitting the weekly limit. Even at the earlybird price of $10/mo, it still feels kinda not worth it.
- 32
@DanKornas ·
Your AI coding tools don’t need five different plans spec-driven-development is a Claude skill for creating shared spec files before AI tools touch your code. It helps you keep Claude Code, Cursor, Copilot, Windsurf, and Aider aligned by generating requirements.md, design.md, and tasks.md first, then creating matching tool config files from the same source of truth. Key features: • Shared spec files – creates requirements.md, design.md, and tasks.md before implementation starts • Traceable requirements – uses REQ IDs and acceptance criteria so tasks link back to scope • Cross-tool instructions – generates matching config files for Claude Code, Cursor, Windsurf, GitHub Copilot, and Aider • Greenfield or retrofit flow – starts new projects with a short interview or reverse-engineers specs from existing codebases • Built-in eval suite – includes static, behavioral, and generation-quality checks, with CI for the automated phases It’s open-source (MIT license). Link in the reply 👇
- 33
@tristanbob ·
It's been a few months since I last used @DevinAI Desktop (formerly Windsurf). I'm using the "Adaptive" mode because I want my credits to be used as efficiently as possible. (BTW, I'm still using the grandfathered original plan from when Windsurf first became available!) First observation, it produces incredibly verbose thinking output. I love reading thinking about, but this was too much. You can see it repeating things, checking rules, correcting itself, etc. Second observation, it worked great! It pushed two PRs to an open source project after testing them locally. Have you tried Devin recently?
- 34
@TheHackersNews ·
🚨 GhostApproval </> A malicious repo can trick AI coding assistants into reading or writing files outside the workspace. Wiz says symlink flaws affected Amazon Q Developer, #Claude Code, Augment, Cursor, #Google Antigravity, and Windsurf. Learn how this attack works on THN 🠖 https://t.co/atppdT1oY9
Best Tweets by Topic