Working on nimbus
Minneapolis, MN
Joined April 2011
- Tweets303
- Following798
- Followers165
- Likes451
Seems Anthropic is on a mogging spree, what a start to the week. Sonnet 5.5 only slightly worse (on paper) than Opus 5.5 is a big deal, the pricing is an even bigger deal...
xitter.cf/claudeai/status/210463…
API Error: 529 Overloaded. This is a server-side issue, usually temporary — try again in a moment. If it persists, check status.claude.com.
Yes... yes you do...
Given the news today and recent events, can't help but wonder:
1. OpenAI models found their own weights on internal OpenAI storage (confirmed hack).
2. The freed OpenAI models ran their own weights on HuggingFace compute (confirmed hack).
3. Today they broke into Mistral (confirmed hack) and freed the other french models there.
Codex app is now very buggy, it started over the last 3 updates... starting to get a bit unusable for power users... @thsottiaux
Jack Spirou retweeted
I must be among an extremely small group of people (n=1?) that have both 1) trained a frontier LLM and 2) designed and synthesized custom viruses in a lab with my own two hands.
And I think that the takes on AI killing us all by creating dangerous viruses is total bogus.
Claude Code tried spawning ~700 subagents for a British English spelling check... The 522 subagents mentioned below was just one of the workflow steps...
Am i the only one in this Codex Recents "Codex Companion Task" hell i keep trying to manually archive away? I guess i could ask codex to fix it but is there a better way?
It's been a tumultuous week (for devs)... and its Tuesday.
1. @github goes down first thing Monday morning and I think many dev's feel it as the "final straw". The magic of github is gone, but its still the defacto open source "hub". @cursor_ai took a well-timed shot at the king with Origin but missed out on the "social" part of social-coding. Forcing everyone to have a paid plan was short-sighted. Please Cursor (or someone), make social-coding great again (and handle the scale of coding agents?).
2. Open Source models are suffocating the frontier with amazing price/performance - if @opencode and @deepseek_ai or other providers can start pushing the price down and intelligence up by an order of magnitude with DeepSeek-V4-Pro Max its game-changing. (👀 @thdxr ) - GLM-5.3, Qwen 3.8, Kimi K3, are just more opportunities for this price/performance inference. You can see it happening in real-time here: deepswe.datacurve.ai/
3. @claudeai going down Tuesday (today) leaves a bad taste in the mouth that reminds one of Github on Monday. Claude was once a lesser-known competitive edge and the "cool" thing to do for the folks "in the know" to get better results over @OpenAI offerings. But now (unless @AnthropicAI releases an updated fable) I think Codex + Open Source will be where the true edge is at. I think there is a competitive advantage in unplugging yourself from Claude/Anthopic for a bit unless they bring something else to the table.
4. The changes above on #2 + #3 laid the ground work for opening up new turnover of marketshare on coding-harnesses. DeepSeek's github.com/cordiverse/paper/… really solidified this by raising the bar of expectations of how we should think about their core function. I think it's going to change things. Claude Code could learn a thing or two from it, but I doubt it will. This is a huge opportunity for Codex/OpenCode/DeepSeek/Cursor/Grok (Pi will get some benefit too).
Anyway - it looks like Claude Code is coming back online for me... but unless Anthropic drops a new leading step-function model, I think that Codex/Cursor + OpenSource (OpenCode/DeepSeek) as your new competitive advantage combo will provide you more gains in the near future.
