George Mitchell

[George Mitchell @mitchellgcc95]

@shaunmmaguire Poch-ball?

[Pietro Schirano @skirano]

You should basically never use Fable for coding, but instead use it as a planner/orchestrator.

Most of today's advanced models can implement a spec perfectly, and once done you can send the work to Fable to review.

This has been my most powerful flow so far.

[George Mitchell @mitchellgcc95]

Fable finally feels like a brain. Using it to orchestrate coding agents is totally different. It seems to really understand & needs no handholding. Great work @AnthropicAI @bcherny

[Crémieux @cremieuxrecueil]

Eli Lilly has done it.

They've gone and made what seems to be a powerful, permanent gene therapy for LDL cholesterol.

That means they'll be able to effectively prevent most heart disease with a single infusion!

[George Mitchell @mitchellgcc95]

@thsottiaux @synthwavedd We love you @thsottiaux

[George Mitchell @mitchellgcc95]

Claude is in a bit of trouble here even if 5.5 isn’t a step change but can finally do Ui & continues to have the massive token usage advantage. If it’s also a step change…..

[George Mitchell @mitchellgcc95]

@ThisWeeknAI I’m in

[George Mitchell @mitchellgcc95]

@Jason I’m in

[George Mitchell @mitchellgcc95]

@fabknowledge Bullish af

[George Mitchell @mitchellgcc95]

@kimmonismus Skill issue

[George Mitchell @mitchellgcc95]

Opus 4.7 is incredible, no idea how people are disliking it. So much better at running through entire jobs, inspecting its own work & iterating. Very good at frontend. More capable 4.6 with the annoying issues seemingly solved — respect @ClaudeDevs @bcherny @trq212

[George Mitchell @mitchellgcc95]

Finally frontend design 4.7 🙌 @ClaudeDevs

[George Mitchell @mitchellgcc95]

@bcherny @EricBuess Thanks Boris

[George Mitchell @mitchellgcc95]

@perplexity_ai @Scobleizer Wow

[George Mitchell @mitchellgcc95]

@iroasmas 😂

[George Mitchell @mitchellgcc95]

@lennysan Skill issue

[George Mitchell @mitchellgcc95]

@Perpetualmaniac I don't think its a bug, I just think they've decided pre IPO to stop subsidising so much to increase value.

[George Mitchell @mitchellgcc95]

@lydiahallie A single terminal agent - 1:15 mins tokens burnt. A month ago it would take about 5 terminal agents & 2-3 hours.

[George Mitchell @mitchellgcc95]

@lydiahallie Yeah, this is such a shame from Anthropic. Working on anything complex is just too token intense to use Claude anymore as tokens burn out in about an hour. Sonnet 4.6 with constant context compact isn’t capable enough.

[George Mitchell @mitchellgcc95]

@RileyRalmuto Totally agree & now if there isn’t a token leniency change it looks like codex is the only viable option at $200 p/m even though a lot of us prefer Claude’s models