Building my own Harness with Claude Code . Much fun , very easy , reliable , customizable , simple . One could call it a >Claw< , its code is just 1000 lines of code as of now . But I think it'll grow a little bit more .

gotta. max out. usage limits. on codex pro and claude max sub. and all the other LLM subs. every week filled with absolute usage

the pentagon just used the same legal designation against anthropic that it uses against huawei and kaspersky . an american AI company , same bucket as foreign state adversaries the department of war offered anthropic a contract for AI in classified environments . anthropic said

swyx put it simply: AI Engineer will be the last job , because they're the ones building the tools that automate everyone else out of existence first anthropic published a labor market study this week and the framing is interesting in a specific way. computer programmers: 75%

claude code now accounts for 4% of all public github commits . 135,000 per day . and a study of 470 open source PRs found AI co-authored code has 2.74x more security vulnerabilities than what humans write alone so we're shipping code faster than ever and the bugs are keeping

i need neuralink to communicate with my agents 24/7 also: when will ai replace government ? looking forward to it .

claude --dangerously-skip-permissions codex --yolo

codex cli gpt-5.4. is ahead of claude code opus 4.6 but opus feels more chaotic neutral vs gpt-5.4. lawful good or something along those lines .

openai sells a model for $0.05 per million tokens and another for $168 . same company , same api , same month . that's a 3,360x price difference two years ago you picked gpt-4 or you didn't . now you're choosing across six dimensions that are all moving at completely different

code output per developer is up 200% . review pipelines can't keep up . so anthropic shipped an AI to review the AI-generated code that landed on monday . claude code review dispatches parallel AI agents to check every PR for logic bugs , not style issues . internally their

three robotics companies came out this week with over $600M in fresh funding , same conclusion: legs are overrated kalanick with Atoms (8 years of stealth , CloudKitchens absorbed into it) , scaringe spinning Mind Robotics out of Rivian at a $2B valuation , and Sunday hitting

websites without needing to scroll - will be great again doomscrolling and scrolling will fall out of favor , other forms of navigation will arise

wait is genpunk a thing now or are you just manifesting

anthropic accidentally exposed close to 3000 unpublished assets through a CMS misconfiguration . buried in there was a draft blog about a model called claude mythos , codename capybara . sits above opus . cyber stocks sold off on the leak 𝗮 𝗹𝗲𝗮𝗸 𝘁𝗵𝗮𝘁 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆

deliberately dissolving influence of the input of the outer world

openai spends 9.6% of its R&D compute on the final training run everyone argues about . epoch just published the split . the other 90% is experimentation , synthetic data , derisking that never makes a blog post less than a quarter of compute goes to the part people actually

This is a dream

Continually observing one's breathing , while feeling the whole body , and approaching inner peace is a cheat code for life

being indoors 80% of your time - recipe for disease imo we are meant to move , to experience nature , to adventure

Your inner compass deserves to be unlocked and employed by yourself