fine-tuning open-source models without this step is just burning... GPU credits. inspect your data first. catch duplicates, empty outputs, bad rows. an AI agent can build the checker for you in minutes. i tried it in Qoder. i gave its desktop agent one task: build a local checker for instruction/output JSONL, let me fix flagged rows, and export the valid ones. on 12 sample rows, it caught: >2 duplicate instructions >2 empty outputs >1 instruction over my 300-character review threshold i fixed one empty output. the valid count went from 7 to 8. Qoder then verified that the exported JSONL contained 8 clean rows. duplicates aren't always wrong, and 300 characters isn't a model limit. those are reasons to inspect a row before training, not blindly delete it. Qoder planned the task, built the app, opened it in the browser, and checked the result. that's a useful tool to have before starting a fine-tuning run. i used Qwen3.8-Flash, which showed 0× Credits in my run. it's free through September 30. Try it through my referral page before downloading Qoder: You’ll get 600 credits: 300 from the 14-day Pro trial and 300 from my code.show more

I MADE MY AI AGENT 10X FASTER WITHOUT CHANGING... THE MODEL not a smarter model, not a bigger context window, not another clever prompt the same kind of AI that designs vaccines for viruses we have not even met yet was spending two minutes opening the wrong files just to hand me a brief from three months ago the problem was never capability, it was the scaffolding that piled up around my agent by accident, folder by folder an agent does not think in your categories, it searches from scratch every single time, and your tidy human folders are a maze to it the fix was almost stupidly small, one index file at the root of each big folder and a few numbers in front of the folder names slowest task dropped from 2 minutes to 26 seconds, fastest ones hit 10, zero model changes capability is cheap when the scaffolding around it is broken the article breaks down the whole system in 15 minutes ↓show more

New open-source agent harness just landed! I got early... access to TrueForge by TrueFoundry and have been running it locally for the past few days. The harness layer deserves as much attention as the model, and open source matters here because you can inspect the loop, run it on your own infrastructure, and swap to the latest or cheaper models. TrueForge handles the runtime work that makes an agent reliable. It drives the tool-calling loop, manages context, coordinates subagents, and executes code in a sandbox, with any model you choose. Every tool call re-sends the growing context to the model, so in practice the harness controls most of what an agent costs to run. A few things stood out from my testing and their published benchmarks. Vendor-Neutral by design. It runs OpenAI, Anthropic, and Google models alongside open-weight models like Kimi, GLM, and DeepSeek. Model routing is a setting, and you can send each task to the model that fits it. On a 14-task enterprise agent benchmark, it matched the accuracy of Claude Managed Agents running the same Opus 4.8 model at roughly 30% lower cost per run (3.8M tokens vs 10M for the same answers). Routing the same tasks to GLM-5.2 held accuracy and brought cost down by about 75%, around $3 per run instead of $12. Fully self-hosted and Open Source (MIT License). I had it running locally with one command, with sandboxed code execution working out of the box. It's time to own your agent harness. Thanks to TrueFoundry for partnering on this post.show more

Own your personal agent, folks. Just built my own... little personal agent with Pi's Pi Durable. I wanted to see how quickly I could build an OpenAI dots clone with open-source tools. I am amazed at how quickly I got this off the ground. It's proactive. It keeps working when you're not watching. It remembers you, runs jobs on a schedule, waits for your approval, and uses its own computer. Pi Durable makes that possible. Every step is checkpointed. I can kill the process mid-task, and My Pi picks up right where it left off. Memory lives in durable documents. What My Pi learns about me in one Space survives a crash and carries over to every other Space. Approvals are durable too. It also has its own computer. Each Space gets a full Linux desktop that My Pi controls with screenshots, mouse, and keyboard. The model runs through OpenRouter, so I can swap models without changing anything else. That's all configurable however you like it. If you've been wanting your own personal agent, now is a great time to build one. I am building this as an experimental ground where I can test new ideas and experiment deeply with proactive personal agents. But I plan to open-source it after testing it a bit and adding more interesting primitives. The goal is to make it customizable and easy to build on, so others can experiment with their own personal agent functionalities.show more

it's f*cking insane Jev sat between GPT and me,... killing every draft that broke my rules before i saw it. good setup. then Jev did not answer. the agent decided silence was safer and stopped sending me anything at all. took a second agent to unstick it. your checker needs a branch for the moment it goes quiet: > let it through tagged unverified and keep moving > hand the risky span to the big model > park it in a queue and retry in a minute > wake a human when the action is irreversible 70-500 ms one Jev call, by TypeSafe's own number. a half-second checker still takes down the whole pipeline when it has no default.show more

Watching all these tech coins gave me an idea,... and I realized it could become something bigger: a real project that people would actually find useful. Welcome to Blueprint. Build fully functional, scalable apps with your own logic, backend, and custom domain. Not just landing pages. Meet Blup, our custom AI agent built specifically for the platform. A complete app typically costs around $1.50 in building credits, depending on complexity. Blup is trained to build high-quality apps that look as good as they work. But your prompt matters too. The clearer your vision and instructions, the better the result. Token holders receive daily free credits based on their holdings. Shortly after launch, we plan to introduce a marketplace where you can sell unused credits. Blueprint takes a 10% share of fees generated by products built on the platform and uses it to buy back and burn the Blueprint token. All project revenue goes toward buying back and burning the Blueprint token. New users get 300 free credits to start building their first app. There’s a queue, so you may need to wait a little. Blueprint GgsSaetRZNCuP9HEWjbojtVWByRWFxpr93S1wtGFpump Read the article: The video shows it in action.show more

🚨Gemini 3.6 Flash is trash I tested it on... a 3D Golden Gate Bridge, and the results were awful. • I had to re-prompt it three times because it repeatedly ignored the instructions. • First attempt, instead of creating the requested .html file, it first tried to build the experience inside the Gemini app using simulations. • Then second attempt it started placing images from the web into the chat rather than actually producing the file. • Even after getting it to complete the task, the final output was dramatically worse than Gemini 3.1 Pro, which is 5 months old and now not even a top 10 model on leaderboards. This feels like a regression from Gemini 3.5 Flash and honestly, it is one of the weakest models I have tested in the past few months. Has anyone else tested Gemini 3.6 Flash yet, and are you seeing the same thing?show more

Google just built Cowork and called it Agent. And... they added one thing Anthropic didn't. You set a goal. It browses the web, digs through your Gmail, checks your Calendar, pulls from Drive then executes the full task. Book a trip. Clear your inbox. Research a market. Done. No back and forth. But here's the part nobody's talking about: There's a toggle "Require a human review." You don't build that unless the plan is to eventually not require it. Google just told you where this ends. I share updates like these in my free AI community on WhatsApp. Join here 👇show more

PART 2: I FOUND 20 MORE PROMPTS IN LAUREN... TAN'S PUBLIC PSTACK. USE THEM IN GROK BOT WHEN "BUILD THIS" STOPS WORKING. part 1 gave you 50 prompts pulled from the public xAI company build. for part 2, I went through Lauren Tan's public pstack, which is available inside her Grok Bot. I converted 20 real playbooks into paste-ready commands for when a build starts touching production. 1. find the failure › 01. reproduce [bug]. find the root cause, fix it, and prove it at runtime. › 02. profile [slow flow]. show the baseline, identify the bottleneck, and measure the result after the fix. › 03. diagnose [live symptom] from instrumentation. do not guess from the source code alone. › 04. inspect [trace]. rank the likely causes and verify the dominant one in the running product. 2. settle the design › 05. /architect settle the caller, types, data shape, and module boundary before writing code. › 06. /blast-radius trace every caller this change could break. prove the condition that makes it safe. › 07. build two throwaway prototypes. assign one agent to each and compare them side by side. › 08. /arena run my request unchanged across multiple models. combine only the strongest parts of each proposal. 3. build against reality › 09. build [feature] behind a feature flag. verify the user-visible behavior inside the real app. › 10. compare [reference screenshot] with [current screenshot]. reproduce the mismatch and iterate until they match. › 11. /tdd reproduce this bug with a failing test first. implement the smallest fix and rerun the test. › 12. migrate every caller to [new API]. preserve existing behavior and delete the legacy path. 4. attack the work › 13. /swarm give one package to each worker. run its check script and return one consolidated report. › 14. /interrogate review this PR. ask several models to find correctness, design, and code-quality failures. › 15. /show-me-your-work keep a reviewable trail of decisions, rejected alternatives, and evidence used. › 16. split [migration] into verifiable phases. make sure every phase ends in a shippable state. 5. finish without babysitting › 17. watch PR 123 until it is merge-ready. resolve conflicts, review threads, and CI failures. › 18. verify every PR independently. report exactly what prevented any PR from merging. › 19. /reflect capture why this task took too long. turn the lesson into a reusable skill or automated check. › 20. /automate-me study my recent workflows and draft a personal mode that routes through pstack. with these prompts, Grok Bot stops being a code generator and becomes an engineering workflow you can inspect. bookmark part 2 for the first task that survives the demo but fails in production.show more

I have a little robot that unlocks my door... from the inside (so I don't have to carry a key) The problem was it wouldn't calibrate on my door. It was designed for 90° turns, but my lock only needs 15° In the past, this would've been a showstopper But I asked AI if it could help... At first, it suggested I snoop on the bluetooth connection to figure out the protocol Then it realized that we could probably just decompile the app's SDK and read its source code. That worked! It told me to switch to my mac and run some python for it. Unfortunately, it couldn't run the scripts directly because Claude Code Desktop didn't have bluetooth permissions, but my shell did. Thinking about it now, I probably should've used Claude Code or Codex in the Terminal. Anyways, it took a while, lots of backs and forths, we had some misunderstandings, it was massively overcautious in the beginning, but we finally got the lock calibrated! Feels like magic! I threw the repo up on github in case anyone in the future (ie me) runs into a similar issue – hopefully it can save them some timeshow more

I BUILT A TOOL TO FIND MY FIRST 100... USERS ON REDDIT 🤯 an ai engine for reddit finds me 50+ potential users and drafted personalized outreach message for each of them all in under 10 mins a day build a guide which tells me which subreddit to post, where and what to comment based on my SaaS just improved the dashboard UI/UX for better thread preview and already crossed 280+ registered users in couple months it get's expensive to run as it's ai heavy tool and not just keyword based unlike other tools but totally worth it just giving away 7 day free trial to all first 300 users, 17 spots left now to try it for freeshow more

✨ I revived my first AI startup from 6... years ago with Claude Code [ 💡 ] Back then it used GPT-3 (this was 2 years before ChatGPT existed!) to generate new startup ideas which then people can vote on And the best startup ideas rise to the top! Back then I made it because people complained they didn't have any ideas to build a startup This week I moved it to its own VPS and installed Claude Code and told it to fix everything, the DB had become big and there was stupid write operations on every page load that it made it very slow Claude Code is excellent at fixing all those small bugs from old projects and quickly fixing them As Garry Tan says "boil the oceans" as in before I'd not have the time to fix these kinds of projects, it wouldn't be worth it, I mean IdeasAI doesn't even make money, but now it takes me an hour to do this and it works again! I also upgraded GPT-3 to xAI's Grok 4.2 for new startup ideasshow more

this is more useful than my entire degree Elon... Musk's rocket company signed a $60,000,000,000 deal for Cursor in June, and eight days ago the two of them put a worker on sale for $200 a month: it gets its own computer in the cloud, signs into your accounts, clicks through your real apps, and hands back finished work instead of a draft for you to paste i ran one against my receipts folder on sunday and got back 14 filed, 2 it held because they needed a card number, and a saved method i never wrote myself Grok Bot is the one you train by doing your own job in front of it, and the whole handover fits in four messages tonight: 1. write out one job you did today the way you would brief a new hire: what has to be finished, which sites and files to work from, what to hand back, and where it stops and asks you 2. let it run once on something safe to get wrong, then correct the result until it is worth your name 3. say "save what we just did as a skill", and add the one rule about what always needs your approval 4. say "run that skill every weekday at 8 and post the result here. if the source is missing, tell me instead of using yesterday's numbers" xAI wrote that order into its own manual: one real job, then the saved method, then the clock. a schedule sitting on top of a method nobody checked replaces two hours of your clicking with two hours of your mistake turns out you never get to pick the brain, and that is the part i would argue about: the manual says there is no model picker for members or admins, no plan to add one, and the bill follows whichever model answered bookmark this, then open the piece below: which jobs deserve a worker of their own, and which ones quietly burn the seat ↓show more

Met my girlfriend's parents for the first time. Her... dad asked what I do for work. I said I build trading systems. He said like Wall Street? I said no. 6 AI agents. They work while I sleep. He laughed. So robots are making you money? I did not argue. I opened my laptop. Showed him the terminal. 6 agents running. 47 mispriced markets caught in the first week alone. His face changed. That is not gambling. That is automation? Exactly. Then I showed him how it works. Built the whole thing in 6 hours. Agent 1: Monitoring Runs 24/7. Watches Polymarket for mispriced markets. Spots an anomaly. Writes to memory and pings me on Telegram instantly. Agent 2: Research Parses news, X, macro data via browser tool on a cron schedule. Every morning I have a full digest on all open positions before I check my phone. Agent 3: Trading Reads the research agent memory. Sees the market has not reacted yet. Acts. Execution tool in gateway mode with a whitelist. No full access on a live server. Agent 4: Watchdog Heartbeat every 5 minutes. Monitoring running. No errors. Positions up to date. Something breaks. Immediate Telegram message. All of this. One Gateway. One config file. Isolation via per-agent scope. The token trick: stopped dumping everything into one file. Critical rules in bootstrap. Markets, patterns, past trades in memory. Semantic search pulls it when needed. Token spend dropped 3x. From $0.40 per request to $0.13. First week running: → 47 mispriced markets caught before Polymarket adjusted → Average entry edge 8 to 12 cents per position → Watchdog fired 3 times and caught a broken RPC before it cost me anything The whole system is plain text files. Open an editor. Change one line. Agent behaves differently. No deploy. No build. Her dad went quiet. Then he asked can you teach this? Her mom asked for the setup guide. I built the entire framework. Six agents. Full deployment. Memory architecture. Telegram alerts. You only need Claude + device + 1 hour per day. Giving this free for 24 hours. To get it: 1. Comment the word "Claude" 2. Like and retweet this 3. Follow me Himanshu Kumar so I can DM you Save this post. Deploy the 6-agent system this week. Start with $200. Scale on evidence.show more

whoever leaked this has bigger balls than sense Google... Research and MIT ran the same agent jobs 260 different ways for Nature last month: they held the prompts, the tools and the compute budget identical and moved nothing but the wiring between the agents, and the same work swung from 70% worse than a single agent to 80.8% better, averaging out at 0.0% i ran my own single agent against the task list first and it cleared 6 of 10 alone, already past the line where a crew starts subtracting this is Graph Engineering, the layer that decides whether a crew is worth 80% more or 70% less, and it installs into the agent you already pay for: - score your solo agent on the real task first: above roughly 45% success that study predicts zero to negative returns from any crew you put around it - under that line, put one supervisor over the fan out: crews with no correction step amplified their own errors to 17.2x the single agent rate, supervised aggregation held it to 4.4x - give every worker one output and let none of them read a peer's draft, so a wrong step reaches the supervisor instead of four other agents - run the comparison again after every model upgrade, because a better model raises your baseline and a higher baseline is what makes a crew stop paying - keep the single agent alive as the control, the only number that says the wiring is earning its calls turns out the shape does not travel: the biggest win came off a finance task under one supervisor and the worst collapse off a planning task with independent agents my position, and it is the arguable one: a crew is a bet on your own diagram, and the model you pick moves that bet less than one arrow does bookmark this, the three moves that draw those arrows before you pay for one extra call are in the post below ↓show more

🤯 THIS FEELS ILLEGAL AND I LOVE IT. I... found a personal AI assistant that actually runs on your computer 24/7 and gets smarter the more you use it. It is called Mercury. 100% FREE. Think of it like having a smart assistant who never forgets anything. Here is what it does: ↳ Remembers everything about you ↳ Asks before doing anything risky It will not run commands or touch your files without permission. You stay in control always. ↳ You can message it from anywhere ↳ Has 31 tools built in ↳ Protects your AI credits One command to install: npx @cosmicstack/mercury-agent 100% OpenSource. Works on Mac, Windows, and Linux.show more

jev + opus 5.5 for design is insane... a... launch video used to mean an agency, weeks of back and forth and a fat invoice. now Opus 5.5 makes the video, the ad cuts and the landing page from one brief. the video below: made in 3 minutes. at 0:09 it drops into edit mode - dot grid, selection handles, pixel sizes on every card - so you see exactly how it was built. but most people get the same mid result: centered text, gradient background, everything fading in. That's Opus's default look when it has nothing to copy. here's the pipeline that makes it look like an agency did it: 1. Build the brand brain first → logo, colors, fonts, screenshots of your real product, your best ad copy → have Opus turn it into a DESIGN.md. The video and the site both follow it 2. Steal a reference, don't describe one → pick 1-2 launch videos with the pacing you want → "make it like this" beats "make it modern and punchy" every time 3. Give it a renderer → install HyperFrames or Remotion: every scene is code, rendered straight to mp4 → a change = one line and a re-render, not starting over 4. Use real components → buttons, cards and UI from 21st_dev instead of empty placeholder boxes → fake-looking UI is the fastest way to look cheap 5. Storyboard before motion → 3 storyboard options, you pick one, then one still per scene → fixing a still takes seconds, fixing a render takes a re-render 6. Run it as a 4-person team + jev → breakdown: pull pacing, type and transitions from the reference → storyboard: the 3 options and the stills → build: every scene in code, music, draft render → review: watch it second by second, list the problems, touch nothing → jev: scores every review note against DESIGN.md (fix now / later / skip) in milliseconds, so build only touches what matters 7. Cut the ads from the same project → same scenes, different hooks: pain first, result first, offer first → 9:16 for reels, 1:1 for feed, 16:9 for YouTube. No reshoots 8. Ship the landing page from the same DESIGN.md → the hero is a looping cut of the launch video → before launch: mobile, every button state, zero placeholder text 9. Give notes like a director → "slow this zoom by half", "hard cut here", "push in on the button" → "make it better" gets random changes paste this into Claude Code 👇 "Read my brand folder and write DESIGN.md. Break down the reference video I give you: pacing, type, transitions. Show me 3 storyboards, then one still per scene for the one I pick. Build every scene in HyperFrames, add music, render a draft. Review it second by second without editing and list the problems. Pass every note to Jev and have it score each one against DESIGN.md: fix now / later / skip. Fix only "fix now". After I approve: cut 3 ad variants with different hooks in 9:16, 1:1 and 16:9, then build the landing page from the same DESIGN.md with the video loop as the hero." everyone has the same model. bookmark this and start using the edge.show more

I did not ask for this and that is... the whole point. Opened my laptop at 8. The brief was already sitting in Slack, timestamped 7:02. Everything that moved overnight, sorted by what it actually costs me to ignore. Three things flagged. Two of them I would have missed until Thursday. Nobody prompted it. No trigger, no "good morning summarise my inbox", no prompt to steal. That is the line between a tool and an employee. A tool waits for you to open it. This one had already done the reading. I still approve everything before it goes anywhere. He proposes, I decide. Viktorshow more

I vibe coded this AfterEffects Animation in 45 minutes.... Step 1: I used Cursor to Generate a type of code called extendscript/java script... No clue what that means, but it worked. Step 2: I ran the script in after effects... ChatGPT voice the whole time guiding me through the process of importing scripts and running it Step 3: I made changes with Composer, just like normal except now after effects is my "localhost:3000" that's just how i think about it in my head. Every time you make changes, you do have to run the script again manually. But I did 7-8 iterations and got this. Step 4: I went to Suno to generate the music. Then elevenlabs to generate the sound effects (wind) and voice. Step 5: Tossed them into premiere, adjusted the audio levels to not be so abrupt... then export.show more

I just vibe-coded a Shopify Reviews Scraper that swipes... all your competitors' customer reviews instantly. 🤯 Paste any competitor's product URL → pull every review in under 30 seconds → export to CSV. Built 100% in Codex. Perfect for brands and agencies who want to mine competitor voice-of-customer data at scale and turn it into winning ad creative. Here's how it works: → Drop a Shopify product URL into the tool → Set your review limit (50, 100, 500+) → Hit scrape — it launches a browser, finds the review widget, pulls every review → Export clean rows: rating, body, author, date → Feed the CSV to Claude and ask "what do customers love and hate about this product" → Turn the output into ad angles, hooks, and headlines No more paying $99/mo for clunky review scraping SaaS that locks you behind credits. What you get: - Any Shopify store's reviews in 30 seconds - Clean structured data (rating, body, author, date) - CSV or JSON export — drops straight into Sheets or an LLM - Unlimited scrapes, no per-review fees - A voice-of-customer firehose you can turn into ad creative on demand This is essentially a competitor review intelligence engine in a box. I'm giving away the full GitHub repo so you can clone it and run it yourself for free. Want the repo? > Like this post > Comment "SCRAPE" And I'll send it over (must be following so I can DM)show more