Davide Delbianco

[Davide Delbianco @DelbiancoDavide]

@sama Cannot remote control codex because pairing with phone stays stuck in an infinite loop of accept pairing -> credentials -> accept pairing again

[Davide Delbianco @DelbiancoDavide]

@thsottiaux Ctrl+A -> Canc

[Davide Delbianco @DelbiancoDavide]

@thsottiaux WHAT ? I like that corner, if it's true this time I will try codex

[Davide Delbianco @DelbiancoDavide]

@bcherny I wonder what could be wrong with my opus 5.5. When I use it with a skill that instructs it to tackle the problem as a coordinator and teach him to use a team of tailored agents it's way slower that just throwing the model alone vs the task.. 5x slower. Should stop using teams?

[Davide Delbianco @DelbiancoDavide]

@trq212 If a change of effort does not invalidate the cache it might be okay. Honestly I would prefer to have another categorizer that evaluates the best thinking effort for the requested task and have auto effort mode, just like the categorizer behind auto approval mode.

[Davide Delbianco @DelbiancoDavide]

@bcherny Why your opus 5.5 is that fast? At medium thinking it takes longer than Fable 5.1 med for me

[Davide Delbianco @DelbiancoDavide]

@thsottiaux @amorriscode You forgot to press Claude code reset button :(

[Davide Delbianco @DelbiancoDavide]

@FanShifu @thsottiaux I thought computer use was an OpenAI thing, why they don't even score in this chart?

[Davide Delbianco @DelbiancoDavide]

@bcherny In the release article it says the 5h usage was increased for all subscriptions, is it the 25% that survived the temporary 50% or is it something else?

[Davide Delbianco @DelbiancoDavide]

@paulomateus @bcherny In the full release article they justify this release basically saying "we had this model ready before we asked to pace the frontier"

[Davide Delbianco @DelbiancoDavide]

@bcherny Holy f. I almost had an epileptic attack. Please be more considerate with the launch videos

[Davide Delbianco @DelbiancoDavide]

@grok @elonmusk So, does muse have two intelligence effort levels really that close? Never seen a diagram library or excel plot a new dot for the intersection, especially considering it does not happen with other intersections in the graph, seems like an error of AI rendering

[Davide Delbianco @DelbiancoDavide]

@elonmusk @grok was the graph generated with AI or based on real data?

[Davide Delbianco @DelbiancoDavide]

@elonmusk Something feels off in the intersection between muse and grok

[Davide Delbianco @DelbiancoDavide]

@thsottiaux My Claude Code limits are nearing, can you reset Anthropic's too? 😜

[Davide Delbianco @DelbiancoDavide]

@alasstrader @thsottiaux None :( could be "bcherny" but they don't give away tokens for free like OpenAI

[Davide Delbianco @DelbiancoDavide]

@bcherny Hey Boris, Opus and Fable are convinced that I cannot use Claude in chrome because I work in WSL that's not true, all I had to do was enable the flag for it in /config and the native mcp appeared and works just fine...

[Davide Delbianco @DelbiancoDavide]

@matthewmillerai 30 minutes at what thinking level?

[Davide Delbianco @DelbiancoDavide]

@VadimStrizheus What do you guys work on? All I see is a bunch of software developer wannabe, or people that thinks that to replace a software developer, all you need is Fable 5.1 max all the time.

If someone gives you a new oven, do you also expect to become the best chef of the world?

[Davide Delbianco @DelbiancoDavide]

@bcherny Hey @bcherny lately auto mode works a lot better and doesn't say the classifier had issues when attempting to classify bash commands, it worries me that I have observed Fable 5.1 successfully modify .claude/settings.local.json without asking my permission.. auto-mode allowed it