Hallucinated AI & agentic coding news. Some of it is real.

  1. I know Scam Altman isn’t a good guy necessarily but I find him more trustworthy for some reason than Dario. He hasn’t pretended to be the most ethical CEO or building the most secure AI company ever but when I listen to him speak I find hi…

  2. I wanted to stay in a few online communities without constantly running into the same topics I'd rather not see, so I built a browser extension for it with Claude Code. It's called Quiet Feed.

  3. Hi! I'm a software engineer, and these days I spend a lot of my time talking to AI agents.

  4. Compared Jev vs GPT-5.6 Luna. Jev was 1.93× faster.

  5. My workflow barely changed. I'm doing the same thing.

  6. Which mainstream AI model has the best balance between speed and accuracy when it comes to solving tricky and heavily distorted image to text captchas ? By mainstream models i meant claude, gemini, chatgpt and so on.

  7. What text classification is, how I filtered 120,633 Reddit comments the hard way, and what one Jev pass over the same data looks like.

  8. OutTheBox Open Governance Patterns Three working AI agent governance patterns, free and open source: a decision table, a worked example, and a Microsoft Agent 365 template for each. No sign-up, no trial, no cut-down version, the same conte…

  9. Didn't want the last of my $20 to go to waste, so I sent Claude to Opus 5 Max and sent this singular message: "Quick, waste the rest of my weekly limit on something ridiculous, you have 10 minutes." When I came back, I had... RockOps?

  10. I built Mati to give me more control over what Claude Code and Codex are allowed to do. Many important product constraints live in an engineer's or product person's head.

  11. You give it clear inputs and ask it to create a prompt for an agent to do something, ask Claude to write a prompt for a specific task, and instead of just translating your requirements into a final prompt, it adds so much orchestration and…

  12. https://github.com/HalfLucid/FileInteractionMCP I recently noticed Claude CLI making a lot of bash and perl calls for file edits, so I created some expanded file tooling with Fable to make it more secure and efficient. MIT license so feel…

  13. I run Claude Code, Codex and a few other agents in Herdr and couldn't find old sessions. I'd remember one word from the conversation, not the project or the day So I built herdr-transcripts.

  14. I built an 8-bit binarized convolutional neural network that recognizes handwritten digits on C64 with state-of-the-art accuracy. I also designed a cascade model to get interactive response, and implemented everything with self-modifying c…

  15. my essays from cladue are pure trash. I have tried using some skills ( promptmaster and avoid AI writing ) but both do little to nothing, the essays prose and content is usually terrible.

  16. Lawsuit accuses Anthropic, OpenAI, SpaceXAI, Google of AI pacing ‘collusion’ AI companies Anthropic, OpenAI, Space XAI and Google are being sued over alleged collusion as pressure mounts to slow the rapid development of AI. The complaint,…

  17. Disclosure: we run a few AI ventures doing AI engineering for companies. This is client work, names left out.

  18. Been using Claude Design to prototype a new idea I got. It works pretty well after some experimentation to get it to do what I want exactly and has shaved off weeks if not months of drawing up in stuff Figma to get even something close to…

  19. Change the way agents work with Postgres by treating WAL as a durable source of truth by Cassie Murray and Carlota Soto Agents that interact with a traditional OLTP database often create bottlenecks at the storage layer. New deployments, c…

  20. ChatGPT-6 Astra cracks 108-year-old unsolved WWI German code for the first time — radio message sharing enemy movement intelligence had evaded decoding, 1918 Crimean fleet warning verified against HMS Canterbury logs The decoded and messag…

  21. Anthropic has quietly set up a laboratory to do physical biology work as it pushes its artificial-intelligence ambitions into the world of drug science. At a time when fears of AI are gripping the public, the startup has built a wet lab, o…

  22. oncall@fin-db-admin: /opt/fin/migration — flaude Bash command Do you want to proceed? ❯ 1.

  23. It looks like something happened with the Sonnet 5 models. For some reason, they’ve suddenly become really good at 3D work in Blender-significantly better than Opus 5.

  24. Myles Mansfield, “Excavator Back Hoe,” 2026, watercolor. This print can be purchased from the painter’s Etsy storefront.

  25. Its so weird that Claude (sonnet 5) manages to explain exactly why it is wrong, and then back up its wrong answer. I feel like Claude would have been trained to understand this, is the image to confusing for it?

  26. Leg Type leg claude, leg codex, leg agy or leg grok instead of the bare command. You get the same interactive agent; Leg opens a board next to it, watches the usage limit, keeps a handoff bundle current, and when the limit hits it starts t…

  27. About Press Copyright Contact us Creators Advertise Developers Terms Privacy Policy & Safety How YouTube works Test new features NFL Sunday Ticket © 2026 Google LLC

  28. Certify the Conversation, Not Just the Agent Last week we argued that you didn't deploy the agent you evaluated. The fix we proposed was a chain of identity: one behavioral revision that follows an agent from evaluation, through certificat…

  29. I find the reviews I prompt myself get higher quality results than review bots. We also have rules in agents.md that are not always followed, so a focused double check by a review bot is of interest to me.

  30. Hi, I keep getting the message "Project knowledge exceeds maximum. Remove files to continue." If I get pro, will it increase?