Garm.com

Claude Code is incredible. The session limits are not. Headroom trims the fat from every input before it ever hits Claude - prompt bloat, boilerplate, verbose docs and more. Same output, half the tokens. V1 out now for macOS → link in comments

Wait, @claudeai gets tired now?

@dadiomov Cards have all sorts of systems in place to STOP robots / humans other than the person whose name is on the cards from spending with them. That makes them a nuisance for agents.

I don't understand the premise that "agents must use stablecoins." Why can't an agent remember the 16 digits of a credit card? Sorry maybe I'm a payments n00b. Stablecoins have a lot of great use cases, but ecommerce shopping is pretty well optimized already for cards.

What an epic first month of running in production :-)

My goal in starting the Headroom OSS project was to make AI more sustainable to work. We have managed to have an active community of passionate users and developers. Today, Headroom hits a milestone. We have saved 100B+ tokens in 25 days! That is: - 350k+ savings - 3M requests compressed @AnthropicAI and @OpenAI continue to build amazing models, and headroom continues to provide the right context to these cc: @claudeai @cursor_ai @openclaw @LangChain @AgnoAgi Try it out: github.com/chopratejas/he…

I kept rocking on

Aww, that's nice

All time productive day so far 👨‍💻

Introducing Origami. chat The world’s first AI that finds you new customers. 1000+ companies use Origami for their outbound. RT + reply with your website and we’ll send you 5 of your perfect customers right now👇

@fin465 Interesting! Curious to try it out with extraheadroom.com

life goals

🏆 𝗠𝗼𝘀𝘁 𝗜𝗻𝗳𝗹𝘂𝗲𝗻𝘁𝗶𝗮𝗹 𝗣𝗮𝗽𝗲𝗿 𝗔𝘄𝗮𝗿𝗱 @ 𝗥𝗘𝗙𝗦𝗤 𝟮𝟬𝟮𝟲 🎉 Back in 2016 I published an article on User Stories together with Fabiano Dalpiaz, Jan Martijn van der Werf and Sjaak Brinkkemper at the REFSQ International Working Conference on Requirements Engineering in Göteborg. At the time, essentially no one in academia was paying attention to User Stories, let alone publishing papers about it - thankfully the REFSQ community welcomed our work with open arms. Every year the REFSQ organizing committee selects the Most Influential Paper from work presented at the conference 10 year prior based on its impact. This year, our paper was selected! “𝘐𝘵 𝘪𝘴 𝘰𝘶𝘳 𝘨𝘳𝘦𝘢𝘵 𝘱𝘭𝘦𝘢𝘴𝘶𝘳𝘦 𝘵𝘰 𝘪𝘯𝘧𝘰𝘳𝘮 𝘺𝘰𝘶 𝘵𝘩𝘢𝘵 𝘺𝘰𝘶𝘳 𝘱𝘢𝘱𝘦𝘳: "𝙏𝙝𝙚 𝙐𝙨𝙚 𝙖𝙣𝙙 𝙀𝙛𝙛𝙚𝙘𝙩𝙞𝙫𝙚𝙣𝙚𝙨𝙨 𝙤𝙛 𝙐𝙨𝙚𝙧 𝙎𝙩𝙤𝙧𝙞𝙚𝙨 𝙞𝙣 𝙋𝙧𝙖𝙘𝙩𝙞𝙘𝙚" 𝘩𝘢𝘴 𝘣𝘦𝘦𝘯 𝘴𝘦𝘭𝘦𝘤𝘵𝘦𝘥 𝘢𝘴 𝘵𝘩𝘦 𝘳𝘦𝘤𝘪𝘱𝘪𝘦𝘯𝘵 𝘰𝘧 𝘵𝘩𝘦 𝘔𝘰𝘴𝘵 𝘐𝘯𝘧𝘭𝘶𝘦𝘯𝘵𝘪𝘢𝘭 𝘗𝘢𝘱𝘦𝘳 𝘈𝘸𝘢𝘳𝘥 𝘧𝘰𝘳 𝘙𝘌𝘍𝘚𝘘 2016. ... 𝘠𝘰𝘶𝘳 𝘸𝘰𝘳𝘬 𝘴𝘵𝘰𝘰𝘥 𝘰𝘶𝘵 𝘤𝘭𝘦𝘢𝘳𝘭𝘺 𝘢𝘯𝘥 𝘶𝘯𝘢𝘯𝘪𝘮𝘰𝘶𝘴𝘭𝘺. 𝘛𝘩𝘳𝘰𝘶𝘨𝘩 𝘢 𝘳𝘪𝘨𝘰𝘳𝘰𝘶𝘴 𝘴𝘵𝘶𝘥𝘺 𝘤𝘰𝘮𝘣𝘪𝘯𝘪𝘯𝘨 𝘢 𝘭𝘢𝘳𝘨𝘦-𝘴𝘤𝘢𝘭𝘦 𝘱𝘳𝘢𝘤𝘵𝘪𝘵𝘪𝘰𝘯𝘦𝘳 𝘴𝘶𝘳𝘷𝘦𝘺 𝘸𝘪𝘵𝘩 𝘪𝘯-𝘥𝘦𝘱𝘵𝘩 𝘪𝘯𝘵𝘦𝘳𝘷𝘪𝘦𝘸𝘴, 𝘺𝘰𝘶 𝘱𝘳𝘰𝘷𝘪𝘥𝘦𝘥 𝘵𝘩𝘦 𝘤𝘰𝘮𝘮𝘶𝘯𝘪𝘵𝘺 𝘸𝘪𝘵𝘩 𝘵𝘩𝘦 𝘧𝘪𝘳𝘴𝘵 𝘴𝘰𝘭𝘪𝘥 𝘦𝘮𝘱𝘪𝘳𝘪𝘤𝘢𝘭 𝘧𝘰𝘶𝘯𝘥𝘢𝘵𝘪𝘰𝘯 𝘧𝘰𝘳 𝘶𝘯𝘥𝘦𝘳𝘴𝘵𝘢𝘯𝘥𝘪𝘯𝘨 𝘩𝘰𝘸 𝘶𝘴𝘦𝘳 𝘴𝘵𝘰𝘳𝘪𝘦𝘴 𝘢𝘳𝘦 𝘶𝘴𝘦𝘥 𝘢𝘯𝘥 𝘸𝘩𝘢𝘵 𝘮𝘢𝘬𝘦𝘴 𝘵𝘩𝘦𝘮 𝘦𝘧𝘧𝘦𝘤𝘵𝘪𝘷𝘦 𝘪𝘯 𝘱𝘳𝘢𝘤𝘵𝘪𝘤𝘦. 𝘐𝘵𝘴 𝘳𝘦𝘢𝘤𝘩 𝘢𝘯𝘥 𝘴𝘶𝘴𝘵𝘢𝘪𝘯𝘦𝘥 𝘤𝘪𝘵𝘢𝘵𝘪𝘰𝘯 𝘪𝘮𝘱𝘢𝘤𝘵 𝘳𝘦𝘧𝘭𝘦𝘤𝘵 𝘩𝘰𝘸 𝘣𝘳𝘰𝘢𝘥𝘭𝘺 𝘢𝘯𝘥 𝘥𝘶𝘳𝘢𝘣𝘭𝘺 𝘪𝘵 𝘩𝘢𝘴 𝘪𝘯𝘧𝘭𝘶𝘦𝘯𝘤𝘦𝘥 𝘣𝘰𝘵𝘩 𝘳𝘦𝘴𝘦𝘢𝘳𝘤𝘩𝘦𝘳𝘴 𝘢𝘯𝘥 𝘱𝘳𝘢𝘤𝘵𝘪𝘵𝘪𝘰𝘯𝘦𝘳𝘴.” In this (ex-)researchers' humble opinion our work initiated a turn around in the field and nowadays User Stories get the attention they deserve, and our paper has been cited 338 times to date in the process! Thank you Maya Daneva, Oscar Pastor, Paola Spoletini and Samuel A. Fricker for the award and the larger Requirements Engineering community for embracing User Stories. This is especially relevant now that the arrival of highly accurate AI is unlocking so much exciting new advancements. I can't to see what the future holds!

Our goal in #headroom was to touch 100B tokens saved in a month - we are 22 days in - and already at 85B! I think we will meet the goal You should join the #discord channel to see how much users #love Headroom! github.com/chopratejas/he… #claudecode #codex #openai #anthropic

@chopra_tejas 's headroom compression engine is doing wonders with claude-code especially after opus 4.7 high token consumption rate. the best things is you won't even feel it's there. It's way better then some token compression YC startup iykyk

@kidAmok @GergelyOrosz Need fewer people on the same team for equal productivity; means those people can work on something else OR you need to communicate with less people in total in your org which reduces error rate

@garmdotcom @GergelyOrosz Why would I want to pay double for double productivity? Aren't I just paying the same rate?

There is massive irony in how AI coding tools are starting to become TOO expensive for many enterprises - after eg Anthropic removed subsidizing AI subscriptions. We might go from "everyone use AI for everything!" to "you have $300/month AI budget; use your brain for the rest."

Opus 4.7 is here. But did you read the fine print? It is using a new tokenizer - What does that mean? - @AnthropicAI charges you by input and output tokens - By using a new tokenizer, they're saying that for the same 'prompt' - you will now have more tokens (so you will pay more!) - Not only that, output tokens, which is the LLM response, is generally more expensive than input tokens, your questions. Agent context is the new infra bottleneck. #claudecode #codex You want to remain unscathed? Try github.com/chopratejas/he… - the only FOSS token compression engine!

@wingedPigBBQ @GergelyOrosz fair enough!

@garmdotcom @GergelyOrosz I am saying it isn’t a small cost. There is no other subscription across eng that is close to that. Budgets are always tight

@wingedPigBBQ @GergelyOrosz How much do you think employing 100 engineers costs? Far above 3.6 million in most places

@garmdotcom @GergelyOrosz That is 3.6 million for a 100 engineers. That is going to be prohibitively expensive for most places. $300 sure isn’t enough though

Glo Dollar is simplifying its structure: the foundation is winding down, and Brale will manage the program going forward. Nothing changes for holders. Glo remains a distinct, mission-driven stablecoin focused on transparency, backing, & impact. Details: glodollar.org/articles/a-new…

@elmorek @GergelyOrosz The better the engineer knows what to do themselves based on shared understanding of product vision, the higher the productivity gained. That + eliminating technical reviews = massive gains.

@garmdotcom @GergelyOrosz In my reality 2x productivity is 100% true. <— would you elaborate? I see it different (20%, maybe a bit more) so would like to understand why you think 100% is common or realistic

@GergelyOrosz 1. sure 2. increased competition among engineers and higher expectations will stop you from doing this In my reality 2x productivity is 100% true. With the latest models and whatever new models are being released in next 6 months 2x is going to be the baseline imo

@garmdotcom 1. as a business owner, nothing stops you from doing so! 2. as an engineer, if AI really makes you 2x more productive, nothing stops you from paying $1,500/month to "outsource" your job 1x and do nothing in the rest Reality is #2 is not true, neither is the 2x productivity...