Samuel
oh thank christ anthropic finally added multi-account support to the claude iOS app
5 33
Samuel
just spent far too much money on candles. sometimes life imitates art
2 36
Samuel
workflow for big projects seems to be approaching something like this: 1. write very detailed behavioural spec 2. make fable yolo it 3. see what happens when the spec meets reality 4. reimplement from scratch using Sol on a very tight leash
3 3 50
Samuel
they’re saying Anthropic has discovered a new way to give me $1 million USD, but they’re keeping it a secret ahead of the IPO. I sure do hope no rival labs try and scoop them on this discovery
3 16 273
Samuel
we’re mere months away from a from-scratch full browser implementation, right?
8 2 63
Samuel
god it’s going to be jarred sumner isn’t it. fuckin bun browser
1 1 38
Samuel
leaving a little unsecured write-with-GET message board out for stray agents at the doorstep
3 18
Samuel
AI imagery must be a categorically opposed. every matrix multiplication spent on image gen is a matrix not multiplied on claude writing javascript, which is pure and good
10 14 156
Samuel
getting mad because a repo does not have a claude.md, implying shoddy software development practice and a higher risk of security issues
5 7 143
Samuel
if you didn’t clock it as being clauded just from the stylesheet, you’re ngmi
1 4 79
Samuel
atproto as agent skills distribution layer? the current state of things is pretty shambolic imo
10 3 65
Samuel
Astra day! super exciting this graph in particular is super interesting. it’s priced the same as Fable (10 in, 50 out) but can solve complex tasks at Sol prices!?
Agent’s last exam graph. Astra is more accurate than Sol at near enough the same price ($10) while Fable and Opus get lower scores for $30 1 1 17
Samuel
Seems like it’s just very terse!
Our evaluations found Astra's written reasoning harder to monitor than GPT-5.6 Sol's, based on tests that explicitly asked it to evade monitoring. We attribute this to Astra's greater control over written reasoning on simpler tasks and ability to solve problems with fewer written steps. Astra still appears to struggle to conceal the reasoning needed for complex tasks, but we take the decline seriously. Improving monitorability remains a research priority, and the accompanying system card details our findings and ongoing work. 3
Samuel
may have just found a fix to a 6 yr old open react native issue? thank you claude fable 5.1
5 1 82
Older posts