Boris shares 2 cases where he uses Claude Cowork:
- status updates from his engineering team, Cowork followus up with the team, gathers feedback and adds it to a spreadsheet
- automatically book hotels & plains based on his accepted google calendar events that are not in SF
From AI @ Scale conference:
- Anthropic works tirelessly to get Fable back to users... - Fable is a massive jump in capabilities, similar to Opus 4.5 vs previous model
- Fable is more nuanced, can get more done
- Boris ran out of hard software problems to throw at the model
Anthropic dropped a ton of great content on their YouTube channel.
Everything you need to learn about AI agents, best practices & real world examples, it's all there.
For some reason Claude Code defaults to Sonnet 4.6 medium even after I explicitly change it to Opus 4.7.
I had to: echo "alias claude='claude --permission-mode auto --model opus --effort xhigh'" >> ~/.zshrc
Looks like Gemini showed me a bit more of their thinking traces, than the usual:
Tone/Style: Balance empathy with candor, validate emotions (if any - this is mostly factual), ground in fact, gently correct misconceptions. Mirror user's tone (straightforward, slightly informal
@romainhuet@romainhuet why is this over reliance on node to run it? I have codex installed via brew, and I don't want node on my mac... It doesn't want to work without node.
I did made my own Claude Code - codex skill where I just pass -> codex exec "command" . Which has no issues vs this:
I start to believe that coding agents have an accidental gamification mechanism built in.
Many more ideas are now low-friction experiments, something that wasn't true about year ago.
You get into a dopamine loop from build stuff, faster than it was ever possible.
One other thing that Anthropic is the undisputed champ [outside of coding] is - the length of their essays... Dario's essay: The Adolescence of Technology - 24k tokens Claude Constitution ~29k tokens