Jon retweeted
Please resist the urge to mock the mathematicians struggling through this moment. Whatever your field, your expertise, your passion, you too shall one day experience what they are going through. And when you remember your words, you will feel remorse.
Jon retweeted
If you didn't feel the ongoing revolution, here is my second project I built with jev.
Your software is drastically outdated. It contains a lot of instructions your CPU executes one by one. Why if we can simply ask Jev for the next instructions to execute
Introducing Jevassembler!
It is a new approach to execute something on CPU. You don't need to wait for code to be written, you don't need to compile anything and YOU DON'T NEED CODE.
You just give it a task and Jev predicts the next instruction to execute. In runtime.
who needs compilers if you have AI
Jon retweeted
SITUATION DETECTED: The US government has sided with OpenAI against the New York Times. The DoJ says training an LLM on copyrighted works does not violate copyright law, and that treating it as infringement would hurt US science, prosperity, and national security.
Jon retweeted
Something is seriously wrong with Claude Opus 5 right now.
I've been digging through recent user reports, and the same problems keep appearing independently:
→ bizarre sentence structure and borderline gibberish
→ unnecessary or invented jargon
→ hallucinating facts, issues or context that were never there
→ correcting itself repeatedly inside the same task
→ ignoring explicit instructions / CLAUDE. md constraints
→ doing work nobody asked it to do
→ creating bugs, then burning huge amounts of time fixing its own bugs
→ declaring tasks finished while leaving major pieces incomplete
→ turning trivial changes into enormous multi-step projects
The communication bug is especially bizarre. People aren't just complaining that it's verbose. They're describing output they genuinely struggle to understand, with mangled phrasing and weird word choices.
Basically, one of the smartest coding models on paper occasionally sounds like a drunk senior engineer who swallowed an enterprise software dictionary.
Anthropic also recorded degraded-performance incidents affecting Opus 5 on Aug. 17, 18 and 19.
Important caveat: that does NOT prove the outages caused these behavioral issues, and I found no evidence that Anthropic intentionally "nerfed" the model.
But the complaints are real, widespread, and remarkably consistent.
Something is off.
@bcherny @AnthropicAI @ClaudeDevs
Jon retweeted
Nocturnal is fully Open Source. I've worked hard for the past 6 months to make Nocturnal the fastest coding agent and personal assistant harness there is. Now you can help make it the fastest growing GitHub projects by giving it a star. Fully open source under the MIT license.
Big moves in this release: desktop companion mode has screen aware dictation and computer use. Basically everything that HeyClicky does and more (backed by Nocturnal's cutting edge agent harness backend and desktop app), but running privately and locally on your machine (totally private if you self host the LLM with ollama etc.). No subscription fee either, naturally. Tested on Linux X11 and Wayland, help with MacOS welcome.
Thanks to everyone who participated in the open beta phase.
If you've been waiting to give Nocturnal a try and have had any frustrations with existing options, especially when it comes to CPU and RAM use, then now is the time to get on board with Nocturnal. Links below.
🤖 Made with AI
Jon retweeted
They're telling you that you shouldn't care about it, they they didn't check if this thing already existed before they vibed it, that they let Claude or Codex make all the decisions for them. When people had the intention to make little effort in the past, they either wouldn't make it or wouldn't publish it. Now someone with the intention to make the bare minimum of effort can publish 100 packages a month seemingly, (in some cases complete domain names and fully generated websites). They could at least have the courtesy of putting a badge on it that says the modern equivalent of "Dictated but not read" (Vibed but not reviewed?), to make it clear that the onus of seeing if it works is on you, dear user, and not the the loop orchestrating mastermind behind the agent behind the project. But at least for the time being these LLM generated READMEs and landing pages stand out like a sore thumb (to anyone in the know at least).
Jon retweeted
Anthropic is rolling out identity verification.
Some users are now being asked for a government issued photo ID and mobile camera verification before they can subscribe to Claude.
Why does Anthropic suddenly need to know exactly who you are?
Jon retweeted
A follow-up thought about this: the "contributor" count to OpenClaw is currently around 2,551. But if you think about it, it's actually closer to the number model/harness combinations used. In other words, you're not going to see the same kind of diversity that you would from 2,551 human developers, since all the models tend to write code in a very similar manner and everyone is using the same AGENTS.md on this project. This fact alone already makes it seem more manageable/predictable than dealing with thousands of humans would be, and probably not wildly different from the result that would obtain from a lone @steipete running the same number of sessions/agents against the project. It's just more people paying for the tokens (minus whatever intent/steering the humans brought to the table).