testycool

@claudeai I was trying to make the most out of Fable today, because I thought it was the last day. I'm at 70% and reset is on July 11th. If I had known I would've used it more moderately.

We're extending access to Claude Fable 5 on all paid plans through July 12.

@_alejandroao The next YT tutorial better include you with the new curls

i got curls --yolo

@simonw As opposed to Opus 4.8

@simonw Also love that about it. It's finally using skills I gave it a month ago, as I had always wanted, that Opus had never used.

After two days with Claude Fable 5 the best way I can describe it is "relentlessly proactive" - here's an example where I dropped in a screenshot of a bug and it span up custom CORS Python servers and used pyobjc-framework-Quartz to capture screenshots simonwillison.net/2026/Jun/11/fa…

@orcdev @krishnasinghdev Do you make a plan with Codex GPT 5.5 xhigh, and then have Claude 4.8 implement it?

really a lot, I plan everything using Codex (GPT 5.5), and do most of my backend and testing with it Claude (Opus 4.8) for executions and UI stuff Cursor (Composer 2.5) for fast executions and less complex tasks although Composer 2.5 can now be used through Grok, so no need for Cursor :)

my AI subscriptions right now - Codex $200 (oss program) - Claude $100 - Grok Build $300 - Cursor $20

@gregpr07 Had the same reaction last year in ~Nov when I tried it again. Since then it's what I use daily for the org I work with, and to unnecessarily automate various things for fun and profit. Also browser-harness is 🫦

yes

@_alejandroao Would also like to know. I've been using Codex since first hearing about possible issues.

will i get banned if i use my claude sub on pi? 🥲

@_alejandroao Can confirm it's good. The previous version got me onto Hermes painlessly. Many thanks 🖤

You can run Hermes Agent on your own VPS for about $15/month. I recorded a quick walkthrough of the full setup: cheap VPS, Minimax token plan, and an AI agent skill that configures the server for you.

@firecrawl Be cool if you actually opened and owned a coffee shop now that I think about it.

Firecrawl is opening a coffee shop in NYC 🔥🗽 On June 4th we're taking over Fellini Coffee for NYC Tech Week, and every single cup is free all day long. Come say hi and grab a coffee on us. RSVP below!

Feeling #blessed that OpenAI finally doesn't ask me to log in + 2FA every time I need to authenticate Codex from another machine.

Claude Dynamic Workflows😩 [user] so...you ran that whole workflow with agents and stuff and did not account for this? [claude] Fair hit. Let me actually check what the workflow said rather than guess. Here's the honest reason it missed it — and it's my fault, not the agents':

@orcdev @alibey_10 gasp

@alibey_10 I don't even need to put everything to notes anymore I just use shoogle.dev

17 new shadcn ui libraries dropped in May 2026 blockus, dominik-ui, toc-cn, gymnopedies, next-labs, bklit, trophy-ui, wensity, corr, paddle, framecn, turbopills-ui, text-ui, amplo, indiacn, cognicatch, stepper added GitHub star count to each registry so you can see popularity at a glance

@patloeber @GoogleAIStudio Same, doing a fitness app with some custom features as we speak. We're all gonna get so jacked 💪

we are not far from everyone building their own hyper-personalized mobile apps. i just vibe coded my own custom todo & fitness tracker in @GoogleAIStudio with zero Android dev experience, and now it’s running on my phone. still feels crazy🤯

@test_tm7873 @Angaisb_ Yeah I should definitely get to setting up some personal benchmarks.

@Angaisb_ tbh agree, also the hype that when any lab like openai or anthropic release anything we all go crazy. 😭like bros, lets wait atleast a hour and test ourselfs if its really good.

Can we stop hyping bad models? Like "OMG look at this 1T model that scores worse than many other open-source models of the same size IT'S SO GOOD" Just stop, really, stop Who's paying you all to post this?

Me switching between tasks all day with AI coding, and life in general

@dbreunig I'm not clear as to why. I've used it on Windows and currently on Linux and it seems fine to me. I gaslit myself into using Bitwarden for a few years because people were saying the UI was ok, some even called it pretty. But no, it was always horrible. 1Pass feels slick af.

1password is excellent when you're on a Mac with TouchID. If you're not, it's slow and constantly breaks your flow.

TIL: You can optimize any agent (cli) with GEPA to automatically optimize your prompts. GEPA accepts any `(str) -> str` callable, it works with your own custom CLI, local models, or API agents. Wrap your agent in a python function and let it self-optimize.

@jamipuchi @chetaslua Yeah, I want to like Gemini 3.5 Flash because the speed is so satisfying. But when it's stupid, it continues being stupid and wastes the quota in 10 minutes.

@chetaslua The reason we don't use codex that much is because it's slower to work with. This would really change the game.

🚨 GPT-5.5-Codex Spark Spotted They hinted at the fact that tomorrow is a HUGE release date 👀 one more thing they are hyping news stuff , tweet was deleted by openai employee check tweet in the comment

@XiaomiMiMo I use the API. Just wanted to say I love both 2.5 MiMo's so much ❤️‍🔥❤️❤️🙏 Fantastic job.

MiMo Token Plan expired? We got you. 🎁 🔹 Expired before May 27 ? We've got you covered — a free Token Plan of the same tier, auto-credited to your account. No renewal needed, no action required. It's just there when you log back in. 🔹 Already renewed after May 27 ? We'll match your renewal amount with balances that go straight toward your API usage. Renewed $50? You get $50 in balance now. Simple as that. Both are valid for one calendar month.

My god MiMo V2.5 Pro is so cool and so cheap I love it so much!