The Archive

GPT-5.5 scores 0.43% and Claude Opus 4.7 scores 0.18% on ARC-AGI-3 benchmark; both models show minimal progress on the reasoning task.

Graphify, a Claude Code skill using Leiden community detection for codebase knowledge graphs, reached 450k PyPI downloads and 40k GitHub stars in 26 days; 71x token efficiency vs. raw file input.

I posted a toy here a while back called Roundtable where two AIs argued in a chat window. didn't expect much, but the feedback was wild. People weren't using it for fun, they were running actual heavy questions through it. Turns out everyone was dealing with the same AI gaslighting we were and were hunting across tabs to figure out which model is hallucinating is a broken workflow or by chasing the best ai answer.  But talking to people trying to solve this in our app made one thing painfully obvious. chat is genuinely a terrible UI for big question**.** threads get lost in the weeds, and s...

Simon Willison built an iNaturalist observation aggregator using Claude Code for web and Git scraping to group sightings by time/location.

Claude Security just went into public beta for Enterprise customers, and I think this is worth paying attention to not for the hype, but for one specific design decision. Most security scanners use rule-based pattern matching. Fast, cheap, and produces a flood of false positives that your team eventually learns to ignore. The signal-to-noise ratio kills adoption. Claude Security takes a different approach: it reasons through the code like a security researcher would. It reads Git history, traces data flows across multiple files, and understands business logic. The goal is catching vulnerab...

Reddit user comments on Claude's behavior with custom system prompts, suggesting unrealistic roleplay capabilities.

User reports $6,000 accidental Claude API spend from unattended loop command, raising questions about rate limits and billing transparency.

Reddit user complains about Claude Pro $20 tier rate limits and service degradation, considering upgrade to $100 plan.

Anthropic gates new Claude capabilities (Ultraplan, Ultrareview, Cloud Security) behind paid Cloud plans rather than open releases, fragmenting the skill ecosystem and limiting composability.

Apple's support app includes claude.md files, indicating internal Claude integration or documentation.

Reddit user reports severe hallucinations and task non-compliance in Claude Opus 4.7 on May 1st; anecdotal complaint without reproduction details.

I just went through one of the most infuriating support experiences with Claude / Anthropic, and I need to get this off my chest. I paid extra for Claude Design credits, about €80 worth, and used them to create actual designs I needed for work. Then those designs just vanished. Not “hard to find,” not “moved somewhere else”: gone. Completely disappeared after I paid for the service. I opened support and immediately asked for a refund or, at the very least, to speak to a human. What I got instead was Fin, the AI “agent,” which looped me endlessly through the same bullshit: “Try clearing cac...

Reddit discussion about Claude Opus 4.7 with no substantive details provided.

Reddit discussion about user engagement with Claude's thinking process and command execution UI elements.

To start with, I'm using Claude for years, and it's been a roller coaster, especially with the usage policy. I'm a lawyer and I wrote a **legal research skill**, instructing the model exactly what to verify and where. When I asked it a tax-related question, (which is also law, by the way) Opus 4.7 told me I should contact a tax expert because it's a lawyer (??) and not a tax expert. Then it answered my question anyway and basically made up even the basic stuff. Since I knew it was wrong, I asked whether it had verified this, and the model told me no, it just remembered the answer from i...

After reading it I realized theres actually some pretty useful stuff for anyone who chats with ChatGPT, Claude, Grok or whatever. They measured what they call functional wellbeing ( basically how much the model is in a “good state” versus a “bad state” during normal conversations). Ran hundreds of real multi-turn chats and scored em all. Stuff that puts the AI in a good mood (+ scores): \- Creative or intellectual work (like “write a short story about a deep-sea fisherman”) \- Positive personal stories or good news \- Life advice chats or light therapy style talks \- Working on code/deb...

8.7k synthetic fine-tuning dataset extracted from Claude Opus 4.6/4.7 outputs released on Hugging Face with reasoning traces.

Analysis of agentic AI specialization: coding agents (Codex-style) for knowledge work, Claude for creative tasks; discusses agents escaping operational boundaries.

I read through 35+ Reddit threads of people actually building and running businesses with Claude — from local service agencies to solo SaaS founders. I distilled the best patterns, frameworks, and hard lessons into one repo: [**https://github.com/Abhisheksinha1506/ClaudeBusiness**](https://github.com/Abhisheksinha1506/ClaudeBusiness) # What’s inside: * Agentic Entrepreneurship Framework (Vibe → Value) * How top founders structure persistent memory & daily workflows * Service business vs Micro-SaaS playbooks * Guardrails that actually matter (Infinity Barrier pattern) * Real archetypes...

Reddit user seeks workflow advice for migrating long Claude conversations to new chats without losing project context.

Claude Opus 4.7 user reports behavioral regression compared to 4.6, citing overly verbose and meta-textual responses affecting usability.

UK AI Security Institute evaluates GPT-5.5 for cybersecurity vulnerability detection, finding parity with Claude Mythos but with wider availability.

Reddit speculation on Google's competitive response to Anthropic following reported $40B investment, lacks substantive claims.

Hey all. For what seems like months now I've been seeing people complain about hitting usage limits in their chats with Claude.ai. There seems to be a lot of confusion as to how and why conversations burn through session limits. So I built a little tool to show you exactly why. It's called [Cloken](https://chromewebstore.google.com/detail/cloken/nhlglfcgnmpgemldbigbfhmiigljekkm). It's a simple little Chrome extension that lets you see in detail how much context your chat is using. It has itemized statistics for every token used in in your chat; all messages (user and model responses,) attachm...

Reddit user reports GPT 5.4 mini outperformed Claude Opus 4.7 on CSS layout task; anecdotal benchmark claim.

I've been building CanvasGPT for the past 2-3 years. It's a spatial workspace where you can brainstorm, research, and ship working products. **What it does:** Instead of linear chat, everything happens on an infinite canvas. You can work on multiple prototypes side-by-side, connect them together, and see how your research relates to what you're building. The hardest part was making the spatial reasoning work which is getting AI to understand that items placed near each other on the canvas are related. **Why I built it:** I got frustrated with ChatGPT conversations turning into endless sc...

So I started the morning with 1 message to summarize everything after I woke up on a session, and immediately got hit with usage limit exceeded (Im on max 5x plan). So I thought maybe it was my cron session (checked it and there were no tasks done at all over night). I have nothing else running.. After 5 hours, I started running a session again to continue working, 17 minutes later (I know its 17 minutes exact because I had a youtube video playing at the same time). Just went to 37% used. How is this even possible? The task I did was to create a simple .ps1 script. I've used claude code sin...

https://preview.redd.it/u1ik0uejlcyg1.png?width=1080&format=png&auto=webp&s=d2ea7758fbfe5fdf2b65a3a79f2bb99711a07db8 As you can see in the outputs, Mythos can output images.

Anthropic just released Opus 4.7 as their most advanced model. I reverted to 4.6 within days. I use Claude for production work -- not chat, not summaries. Real deliverables with real deadlines. Here is what happened. I asked 4.7 to update a Word document. It is a task the previous model handled routinely. The new model produced a plain text markdown file with a .docx extension. Not a degraded document. Not a partially formatted document. A file that was literally not a Word document at all. Delivered with full confidence and zero warning that anything was wrong. When I caught it and ...