jack trader

@gregalthoff @abuchanlife @Fried_rice bros the one who posted it

@abuchanlife @optic0n @Fried_rice Dude.

Claude code source code has been leaked via a map file in their npm registry! Code: …a8527898604c1bbb12468b1581d95e.r2.dev/src.zip

@recap_david have you actually used it yet lol

I don't think people understand what this actually means. Every application on earth can now build an agent that teaches ITSELF how to use the application through the UI. Not through API integrations. Not through documentation. Through the actual interface, the same way a human learns. Here's the loop: You define what success looks like (an eval). You point Claude at your application via Computer usage. Claude tries to complete the task through the UI. It fails. It writes what it learned to a skill file. It tries again. Recursively. Hundreds of times. This is Karpathy's auto-research method applied to software usage. Let me make this concrete. I built a company called CoinLedger — crypto tax software, ~1 million users. The product is powerful but complicated. Users have to import wallets, classify transactions, handle edge cases, and generate accurate tax reports. The learning curve is our single biggest challenge. With Claude computer use, I can hand it public wallet addresses and CSV files and say: use CoinLedger to produce an accurate capital gains report with no errors. Claude opens the app. Navigates the import flow. Hits an error. Documents the failure. Adjusts. Tries again. Each cycle produces better skill files. Each skill file captures how to properly use a specific part of the app. After enough iterations, Claude has built a complete agent harness — a set of instructions that lets it use CoinLedger as well as our best power user. Then I ship that agent to every user who struggles with the platform. The biggest friction in a million-user product, solved by an AI that grinded through the learning curve so humans don't have to. Now multiply this across every complex application. Every SaaS product with a steep onboarding curve. Every enterprise tool where 90% of users touch 10% of features. The first applications that build these recursive agent harnesses will compound in ways their competitors can't catch.

@Retsevi @Ritikachoudhar coed gym bathrooms are great

@Ritikachoudhar Nobody is fucking at the gym

A lot of married women are getting fucked at the gym.

@Ritikachoudhar confirmed. my ex was getting railed by chad in the family minivan. literally.

@sandislonjsak they’re shipping slop.

My brain simply can't run more than 3 agents in parallel and QA all of their work. I am sure I am not the only one. How do people manage 10 at once? Or they simply lie?

@meetCalAI @altryne no

@altryne Prompt caching is the unsung hero of the agentic era. If you aren't obsessing over context management, you're burning venture capital for warmth. How are you handling cache-invalidation at scale without over-engineering the state machine?

PSA: If you've been running out of Claude session quotas on Max tier, you're not alone. Read this. Some insane Redditor reverse engineered the Claude binaries with MITM to find 2 bugs that could have caused cache-invalidation. Tokens that aren't cached are 10x-20x more expensive and are killing your quota. If you're using your API keys with Claude this is even worse. This is also likely why this isn't uniform, while over 500 folks replied to me and said "me too", many (including me) didn't see this issue. There are 2 issues that are compounded here (per Redditor, I haven't independently confirmed this) : 1s bug he found is a string replacement bug in bun that invalidates cache. Apparently this has to do with the custom @bunjavascript binary that ships with standalone Claude CLI. The workaround there is to use Claude with `npx @anthropic-ai/claude-code` 2nd bug is worse, he claims that --resume always breaks cache. And there doesn't seem to be a workaround there, except pinning to a very old version (that will miss on tons of features) This bug is also documented on Github and confirmed by other folks. I won't entertain the conspiracy theories there that Anthropic "chooses" to ignore these bugs because it gets them more $$$, they are actively benefiting from everyone hitting as much cached tokens as possible, so this is absolutely a great find and it does align with my thoughts earlier. The very sudden spike in reporting for this, the non-uniform nature (some folks are completely fine, some folks are hitting quotas after saying "hey") definitely points to a bug. cc @trq212 @bcherny @_catwu for visibility in case this helps all of us.

@LukasHozda 🤣🤣

Garry, buddy, for loops exist you don't have to repeat the same statement 37k times for your blog

@cryptopunk7213 in todays episode of "x replaced software engineers" >

well thats fucking it - anthropic has officially replaced software engineers. claude is now a 24 hr autonomous coding agent. claude can now operate your entire computer and CLAUDE CODE = end-to-end software engineering: - claude writes the code for you - then literally opens the app it coded - clicks through the entire app and find bugs - then fixes the bugs and improves the app in hours. previously claude generated code, you run it and give claude feedback. thats completely gone now. all in a continuous loop without leaving your terminal 😂 we're barely through monday. well done lol

would really love to see a trace after being limited like this to be able to surface what’s causing the most token usage. show the user a breakdown of why their token usage is so high for a given period. maybe they’re doing things they didn’t realize that seem simple to them but are extremely costly on the backend. xray traces but for token utilization 👀

Working as hard as we can to make this better. It’s not easy growing at this rate, and it has been straining our services. Thanks for bearing with us. A number of significant improvements landed in the last few Claude Code releases and a few more on the way. Make sure you’re on the latest version.

Claude code is essentially unusable for me i hit 36% session limits in 15 mins. Just by using a single agent on my codebase. @claudeai you need to fix it. I'm on the 100$ plan btw

@acxtrila 🤣

How tf are you adding 78k more LOC to a newsletter website

as an SRE who spends most of his time making failures visible to my teams, i’m always trying to make debug information self service. is there a plan to enable historic usage tracing? for all these people saying they’ve hit their limit at 8am, boy would it be awesome to have a trace output from claude code to show you exactly what’s eating up all their tokens at an account level. i’m sure this would save yall a ton of headache and support churn. 👀 @trq212

Hope this was useful! I wanted to keep going but had to stop myself. Will post more soon. What are your favorite underrated Claude Code features?

I wanted to share a bunch of my favorite hidden and under-utilized features in Claude Code. I'll focus on the ones I use the most. Here goes.

@k1rallik what have you built with it so far?

> been paying $200/month for cloud AI APIs > laptop: M2 MacBook, 16GB RAM > tried running models locally, garbage quality after 4K tokens > read this TurboQuant breakdown on Tuesday > applied 3-bit KV cache compression > same MacBook now runs 100K token conversations > quality: identical to cloud > cancelled all API subscriptions Wednesday > it's been 3 days > saved $200/month forever > with a free algorithm from a free paper > my MacBook didn't change. the math did

@MoeCanDoIt skill issue.

I'm on Claude's $200 plan, and for some reason it's getting dumber. I'm talking simple contract review stuff where it COMPLETELY misses important details. Did it get nerfed? What's happening?

@trq212 i know yall have been shipping a ton but im pretty sure you need a reliability and performance pass. its getting really bad. agents and sessions just randomly die with no errors, tool calls just fizzle out seemingly randomly, remote sessions are unreliable, the list goes on..

highly recommended workflow here. i started adopting something very similar using the superpowers method at work and on personal projects - the results are significantly better. my problem now comes reviewing code from others who don’t use the same process. i’d love to see the spec alongside the PR to see if you’re actually solving the problem you set out to solve. curious how you handle this.

I'm not 100% happy with the quality of code claude or codex generate. Theres a lot I discard or go back and finesse the low level architecture. Also at work I'm seeing too many problems arising out of people abdicating critical taste and judgement calls to the LLMs. But that said my extensive use of agentic engineering both at work and personal side projects has brought a lot of learnings on how to effectively write code via agents. This is a collection of those learnings.

@adamghaida agentic availability

jesus what is happening

@trq212 are you hiring SRE’s?

To manage growing demand for Claude we're adjusting our 5 hour session limits for free/Pro/Max subs during peak hours. Your weekly limits remain unchanged. During weekdays between 5am–11am PT / 1pm–7pm GMT, you'll move through your 5-hour session limits faster than before.

@shreypandya @browserbase nobody could predict what will happen here. /s

Introducing /cookie-sync Run browser tasks in the cloud with all your authenticated accounts, powered by @browserbase Watch as my agent: - uploads my local Chrome cookies - injects them into a remote browser - goes to Forkable & chooses my Friday lunch for me

@rezoundous supervised AI yes.

So it has been about a year since. Is AI writing 100% of your code?