James

@thdxr roughly 3.8B tokens ago probably

at what point should i check in to see what's going on

@awscloud How it started --> x.com/i/status/20773…

damn @awscloud honestly it's not the 26 Billion Dollar Bill that's suprising.. but that you couldnt find *any* Savings Opportunities💀 #ngfmi

@thsottiaux please treat us like adults and never bring back the 5hr limit again.

We've had no 5h limit in Codex plus and pro for a few days. Do you think it is better or are you finding it difficult to manage the usage included in the weekly limit effectively? If we were to make this different, what should it look like in an ideal world?

@ChatGPTapp and 5.6 Terra has been a rock. It just spun up and optimized a stable of fine-tuned open weight models on an aws instance for me. Costs less than $3/hr for 96GB RAM and 3.5TB of storage.

@arvidkahl Try @MulticaAI It is Insanely fucking cool and easy to use. I have Opus coordinate it all through the CLI - tell dem bishes "no worktrees, TRUNK-based only" works like a charm multica.ai

I want to run Clauded Code AND Codex on the same codebase at the same time, but I really don't want to have to deal with worktrees and provisioning db/redis/opensearch for each of them...

@trikcode every-damn-day lol >copy the convo thread in .md >save to ./sessions >spend a few minutes deleting screenshots (hundreds of lines of nothingness) >Fresh session "please get familiar. last session is at ./sessions" >resume flow state ✨

Got rate limited mid-task yesterday and had to switch from Fable to GPT-5.6 Terra. 6 hours of back-and-forth context entirely gone.Lost 90 minutes just getting a different model up to speed on the project goals. Hitting a usage cap at the worst possible moment sucks. We desperately need a universal way to share context across different models.

The psyche of using a "less-capable" model [for any type of task] is a serious mental blocker and heavy financial burden for me. @claudeai Fable be Toooken.. smoked 30% of 10k credits in less than 2 weeks. Thanks for the 1-month token-spree @GitHubEnt @github Love ya💜

@leodev did you try domains-sdk.. compromise but shit.. bascially the same thing.

Just perfect. Some random guy is sitting with the package name "domain-sdk" and it's not even a real thing. I don't even know what to do atp... like what's the point of contact npm if they won't respond or won't care about this 😭

@domanski_ai hell yea! love to see it. If i could afford Claude subs I'd be using headroom right now 😅

@jamesnavinhill That's basically headroom's routing today. Preference order, skip anything that's hit its limit. The <10% reserve floor is a sharp tweak and I'm adding it. Thanks bro!

"You've hit your usage limit." Then you burn ten minutes logging in and out of your other accounts trying to find one with anything left in it. I got tired of that and built headroom. Every Claude and Codex account you own, on one live dashboard. It reads all of them without spending a single token. The moment one hits a wall, it rotates you to the next account that still has room. Free. MIT. How many accounts are you juggling right now?

@grimcodes absolutely gorgeous!!

Can you really call dither-kit a KIT when it only contains charts? introducing dither-kit primitives: - dithered buttons - dithered avarars (with up to 1.5T variations) - dithered gradients (the coolest if you ask me) try them out with: npx dither-kit/latest add

@theishanvyas boom! please bless us with your tokens 🙏 Working on an open-source platform with provider-agnostic adapters at every layer. Your open-weight harness would do wonders for our development. Runnign final tests now, launch imminent. github.com/studio-jami jami.studio

*Unlimited coding tokens for a year.* I am giving 100 strong builders access to a coding harness built on top of 20+ open-weight models: GLM-5.2, Kimi K2.7, MiniMax M3, Qwen3.6, and more. All unlimited. Want in? Reply with your GitHub + live project. Then DM me: “open source” Picking the best 100.

@marty_kausas Working on it. Jami.studio is provider/ host agnostic and has curated workspaces for business, project mgmt, design, research and coding. Installed to your preference. Don't mind the rough landing page, the product is sharp. Will be ready in a week. Final testing now

i'm so sick of using claude code in a terminal i'm not coding. who has made a great app that i can use with any model?

@AlexFinn I often color it up with something like "DO NOT TASK THE USER. This isnt a fucking Easter egg hunt." seems to work well

My favorite ChatGPT 5.6 Codex tip: “Just do it yourself” Anytime I’m building in Codex and it asks me to do something (log in, download this, click on that, get an API key, sign up for this) I just respond “just do it yourself” Codex is SUCH a good agent harness it literally 100% of the time can do all those tasks itself Much like yourself, it is constantly underestimating its own abilities. It’s computer and browser use is so good that basically no task online can’t be automated by Codex now What you’ll find by doing this exercise is basically 100% of your knowledge work can be automated All the manual steps that go into building things can now be done by AI agents. Bonus tip: after it does something it asked you to do, tell it to turn what it did into a skill so that moving forward it never asks you to do it again If you do this over and over, you’ll end up saving SO much time in the long run because Codex will be able to automate basically everything you do on a computer Start implementing this immediately

~ Benefit Hunter ~ Helped me lock in over $200k of development tools for every layer of the stack. For quick apply or tracking opportunities: benefits.jami.studio Repo-starter with setup, example applications, and scripts: github.com/studio-jami/be…

~ Do Not Disturb ~

@ivan_bezdomny @ClementDelangue Really nicely done! Signed up. Fyi: the header on mobile is causing the page to extend width-wise both for signed out/in [galaxy s24 ultra- chrome mobile] Easy fix. big win. Keep it up

@ClementDelangue Wrote a little X article on how we build HuggingNews here. Especially how we trained an open weights model to do the final writing. x.com/ivan_bezdomny/…

Keeping up with AI news is becoming a full-time job. So my friend @ivan_bezdomny built HuggingNews, an AI-curated feed that surfaces the news actually worth reading. Soon, it will even personalize the feed using your Hugging Face profile. Been using it for weeks! Bookmark it, or ask your agent to send you the top 10 stories every morning or night. Less noise. More signal. More building! huggingnews.com

@dwr @getpieces will do that and more. Set and forget easy. @builderio has an open source agent-native repo with a lot of incredible apps including a screen recorder/session-replay built into their Clips app. github.com/BuilderIO/agen…

Is there a cost effective and straightforward way to: 1. Record video of everything I do on my Mac for a week 2. Feed into an AI 3. Have the AI recommend concrete things to automate? Coach me on stuff I'm inefficient at?

@define_app ohh that looks delicious, ill bite.. define.app/card/019f44f3-…

The `/human-taste` skill is crazy good...