martin.p

we shipped fat man walking today habit tracker where your character literally evolves from couch potato to beast as you build streaks.. the character on screen gets fitter as you do live on the App Store now

@louisvarge sounds like it can easily end in an endless loop? any max iteration or way for agents to abort?

i made a thing where now any Claude Code can send messages to any other Claude Code on my machine they can ask clarifying questions about work, or become friends

now we're talking, 34% of claude max used within a day. running 3-4 agents with /loop through the night, each spawning their own sub agents. can I finally use the full 100%?

@noahzweben @bcherny does this run via my claude subscription or API credits required?

You can now schedule recurring cloud-based tasks on Claude Code. Set a repo (or repos), a schedule, and a prompt. Claude runs it via cloud infra on your schedule, so you don’t need to keep Claude Code running on your local machine.

@trikcode too real, it's actually over tho

There’s a new kind of burnout now. Not from working too much. From trying to keep up with tools, models, frameworks, launches, and 600 people saying “it’s over” every morning.

quick hack on how to run /loop within claude code when you want it to start later because it requires another terminal's loop to finish beforehand :D gotta make the ai work while you sleep

great and long week of building in Japan ends with an amazing Asahi Super Dry beer and chill. First time looking at the real possibility of my AI subscriptions bringing in more money than they cost 😂

@shiri_shh it's literally cursor bench. it's certainly a great achievment, but it does not mean composer 2 > claude, as it doesn't compare opus in claude code vs composer 2 in cursor

Composer 2 just beat Claude Opus 4.6 and it costs almost 10x less yeah… read that again. > A 50-person team built their own model from scratch and then outperformed one of the most funded AI labs on coding… >61.7% vs 58% on terminal-bench. this is not normal. and it’s not just benchmarks > they fixed the biggest pain in AI coding too you know when your AI just forgets everything mid-session? yeah… that’s gone too Composer 2 keeps summarizing its own work so it doesn’t lose context in long coding runs when a small team can ship something better, faster, and 10x cheaper… this whole “only big labs can win” narrative just broke..

while that's an amazing achievement, I do assume that this benchmark only compares opus IN cursor vs composer in cursor. so here composer is certainly highly specialised for tool usage and orchestration, while claude is not. a really interesting benchmark would be to compare composer in cursor VS claude in claude code. if composer comes out on top there, I'm back in the cursor business.

🚨 Cursor just dropped Composer 2.. their own AI model.. not Claude.. not GPT.. their own.. and it beats Claude Opus on coding benchmarks.. at a fraction of the cost.. a code editor with 50 people just outperformed a $30 billion AI lab.. at coding.. which is supposed to be their whole thing.. the vibe coding era just got an upgrade..

setting up a well built skills library for claude code is paying off big time right now - actually feels like magic

@levelsio just had the same over here, it worked after an hour or so, I think the servers were just actually down or smth

Every week at least one day I am stuck in some weird Claude Code logging me out somewhere and then their OAuth is completely broken and I get stuck and can't code

@techhalla would love to beta test if u need someone :)

@OnePromptMagic Not yet but dime of you asked for it so I'll do my best

The game engine I'm cooking (RAMEN) is finally coming together! 🍜 Anyone down to test it out? I'm droppin the first version next week and it's gonna include: · Video to spritesheet generator (AI video compatible) · Spatial detection for 2D backgrounds for custom lights and shadows · Node-based game logic. Totally no-code · Tons of other stuff focused on making a 100% playable 2D adventure in just a couple of days This is just the tipo of the iceberg, so saty tuned for more details, but if you're intersted on testing it, just let me kno in the comments. Thanks!!

a yer ago I would have said, start learning any programming language without AI to get some basic understanding, then gradually introduce AI into your workflow.. nowadays with the models getting ridiculously good, I'd say start right away with AI, but learn some basics on security (RLS policies, api endpoint authentication, rate limitting etc)

Which is best for beginners: 1) VS Code 2) Cursor 3) Vim 4) Antigravity

@DataChaz @AnthropicAI sadly first thing they ask you is if you belong to the partner network or not. if not they might remove you from the program

🚨 There’s a new gold standard if you want to build multi-agent systems and enterprise tools. @AnthropicAI just dropped the `Claude Certified Architect` certification. It’s a rigorous 60-question, 120-minute exam that proves you can build production-grade applications. To pass, you need deep, applied expertise in: → Designing agentic loops & subagent patterns → Building custom Model Context Protocol (MCP) servers → Configuring CLAUDE.md hierarchies for CI/CD → Forcing structured output with strict JSON schemas It's rolling out at $99, but early access is completely FREE for the first 5,000 partner company employees. Certification link in 🧵↓

@aditiitwt it's insane, like 90% of the comments here are clearly AI slop 😂

Imagine your API scaling from 1K → 100K users. Suddenly it starts failing. You’re not allowed to upgrade infra. What tricks would you use?

@Manixh02 I'm not applying for a backend developer position, but for an agent orchestration position with the edge that I have real world knowledge in backend development. With new tools and models coming up, you'll need only one or two good ai tool users, and I can be one of them.

Backend Developer interview: We can write 80% of our code with AI why should we still hire you. What will be your response?

at some point AI will build flawless code, and then it won't matter anymore. for now yes, this can be frustrating (even as software dev). One cool thing I noticed is, stuff I tried to vibe 2 years ago as experiment for testing boundaries and got stuck soon, I now retried and it worked better than what I could have hoped for. My flow: rename the old vibed repo to 'old', create empty folder 'new' let Lord Opus rewrite the app in whatever other programming language -> it won't be able to copy stuff over but has to 'rethink' the process and it's just gonna be a lot better like this. At least that's my experience :)

Vibe coding is so easy. Until you don’t know how to code and you’re stuck with: - runtime errors - broken logic - AI hallucinations - messy workflow - laggy redirects and you have no idea how to fix them.

@jackfriks sounds pretty tbh :D can you claim a refund?

i really thought deep and hard and almost worked up enough courage to spend a stupid amount of money on an apple studio display only to realize it’s 60hz 😹😹😹😹😹…

@VadimStrizheus Go for a walk :D

CLAUDE IS CRASHING AND NOT WORKING FOR ME what am I supposed to do?

first real outreach campaign for inboxmate about to start. we got the leads, the pipeline, mail system setup with initial and followup. individualised demo for every lead setup and viewable without signup. claude code drafts every email personalised based on crm activity (refers to previous emails) and lead website content. I setup a service with its own mcp that allows claude to create only the email drafts (with nice looking templates), but NOT be able to send them (we don't want to go full yolo here). the system looks good, but honestly, this is mostly an experiment for fully agentic led marketing.