stewones

@jarredsumner Your reasoning aligns with semver but idk man, it deserves a v2. Would be much clearer for consumers wanting to adopt it.

We’ll call the next release Bun v1.4. There’s already far more than enough new things since Bun v1.3.1 to justify the version bump, and the slightly more opt-in upgrade of a minor bump seems better.

Still writing blog post hopefully will post tomorrow

@dschewchenko @saltyAom cold starts

@saltyAom You don't need to run it on a 50 year old computer, which is so important to have a very compact program. Why is it so important to have the smallest bundle for the server ?

The bare-bone server brings memory close to Hono The bundle size is not too far Elysia 2α: ~57kb Hono: ~20kb Note that Elysia has a JIT "compiler" and Static Analysis, larger fix cost but faster Elysia 2α also uses less memory as app scales *Bundle size is not a final number

@VictorTaelin have you tried a combo of Skills + Harness’ hooks. it sounds like a great usecase for your domain.

seriously, working with AI is MISERABLE for one and only one reason: having to re-explain the same thing "oh yeah this new session obviously doesn't know what proper case trees are, so let me explain it for the 5000th time in my life" I'm tired AGENTS.md doesn't solve this because it is impossible to fit the entire domain knowledge without nuking the context - it would be 1m+ tokens worth RAGs don't solve this, the agent won't search unknown unknowns SKILLs don't solve this unless I keep like a collection of 1750 skills with specific cuts of domain knowledge for each possible subset of my domain that I might need in a given chat, but that's a lot of manual work recursive LLMs or whatever don't solve this for the same reason, you can't dump a domain book and expect the AGENT will magically guess that it is supposed to search for a specific bit knowledge. unknown unknowns fine tuning doesn't solve this (OSS models suck and OpenAI / Anthropic gave up on user fine tuning) I honestly think a good product around fine tuning on your domain would be a major hit and an underdog lab should take this opportunity

@dmythro @zeddotdev Newton’s 3rd law.

@zeddotdev That’s rude.

We've shipped more than a thousand versions of Zed, but all of them began with zero. Today, that changes. zed.dev/blog/zed-1-0

@_glnarayanan @themkmaker that's exactly what I needed, thanks for sharing dude 👍

@themkmaker I paid for it to support Adam & Tailwind. Initial experience, it's very similar to Impeccable.style.

Got the ui.sh invite today. But the page doesn't tell anything about what it offers for $120/yr. Anyone already tried it?

@xdNiBoR i’m from brazil seeing it in english, using the app. so i’d say it works 🙃

Nu dat X posts automatisch vertaald, kan ik gewoon in het Nederlands posten. Ik vraag me af of nederlandse posts evenveel reach hebben. Zien jullie dit in het Engels?

@JuanRezzio @cursor_ai nice but you just recreated the codex app. and judging by the economic aspect, we’re never leaving openai. you guys really have to work on making your pricing and limits more attractive.

Last thursday we released @cursor_ai 3.0! Are you guys liking it? Is it your go-to coding layout or are you using the IDE? Curious about your thoughts!

@samuelrizzondev Tarefa -> Refinamento -> Planejamento -> Breakdown -> Fatias verticais 100% testaveis (max 1k loc) -> Review humano

Tô com um problema muito grande no trampo no processo de code review com IA. Estamos trabalhando com épicos, o que significa que os PR’s são maiores, então a gente coloca a IA pra revisar. Mas o problema é que a IA SEMPRE sugere mudanças, e não são poucas. E como são muitas sugestões, usamos IA para corrigir as sugestões. E agora estamos em um loop infinito e a única forma de resolver isso é fazendo um review manual, mas como são muitas coisas tá bem difícil. Se tiverem alguma sugestão eu aceito, pq esses dias passei um dia inteiro fazendo code review de uns 10 PR’s

@BonaVictor @brunobertolini das features que mencionou, a única que faria diferença é o docker-compose. foca nisso.

Eu entendo totalmente oque você quer dizer, mas é por isso que eu trouxe pra plataforma fluxos iguais ou semelhantes a aqueles que a cloudFlare e Vercel ja tem! Por exemplo, de tudo que você mencionou, só não temos web view analytics nativo e preview env! Ambos estão no meu roadmap entretanto! Você literalmente conecta seu github, cria um serviço e nós cuidamos de todo o resto! Mas eu entendo, é complicado quebrar a barreira dos grandes players! Por isso estou focando em trazer valor com features auxiliares, como por exemplo: - Catalogo de serviços, você vai poder fazer deploy de varias aplicações com um unico click, tipo redis, postgres, mysql, valkey e etc. - Um bucket storage nativo 100% compatível com s3! - RBAC completo. - AI Gateway nativo PAYG. - Suporte a docker-compose. Você acha que com valor agregado, seria possível quebrar a barreira do conforto dos big players? Pergunto honestamente, pra mim poder entender pra onde preciso ir!

Uns meses atrás, eu vi alguem compartilhando um micro-saas, de 1 feature, que bateu 50k de faturamento em pouco tempo, e pensei: cara, vou tentar. Ontem fez exatos 60 dias do lançamento do meu primeiro micro-saas, e faltou pouco pra bater 50k de faturamento (feriado atrapalhou). 50k em 60 dias, com um micro-saas, feito 100% com IA em 3 dias, 3 dias cara!

@steipete @builtwithjon @builtwithjules if you want a full agentic lifecycle you def need a plan mode. and that doesn’t stop at planning, you also need agents to refine and breakdown tasks, map dependencies and execute them in parallel. that’s a completely different paradigm.

@builtwithjon @builtwithjules i do the same. Just doesn’t need a “mode”. Most replies here don’t get that

I always plan. What's your workflow look like? I talk through what I want and get it to a specificity suitable for an agent to implement (or a human for that matter). More about figuring out what's needed than instructing the agent. Helps to have @builtwithjules ready to go!

@ThePrimeagen @javilopen I also maintain my SKILLS.md every morning for 1-3 minutes

@javilopen I do it every day, every morning for 1-3 hours Because I want to maintain skills

Serious question: Is there any programmer left in the room still coding the traditional way, character by character, without using AI? If so, why? Explain your reasoning.

@fredrikalindh @itspers 🤯

@stewones @itspers it does it on its own - no need to do anything

reviewing in cursor is now a much better experience than github - select diff and ask cursor why it's there (or to fix) - view videos/images of result - test straight from browser we also added mark as viewed, link to preview and many more improvements coming

@fredrikalindh @itspers how?

@itspers it can do this already!! the agent can use subagents and choose faster, cheaper models

@LukeberryPi Simples, windows/linux users não gastam dinheiro com software.

>apenas pra macOS pq as grandes empresas não tão construindo produtos pra windows/linux?

@sacsbrainz @jarredsumner @kalomaze hell yeah, just figured that yesterday. thanks buddy!

@stewones @jarredsumner @kalomaze cd into your project dir, open Claude Code, run /resume to list past sessions. Sessions are per project. Use /resume-session for your last convo globally.

wtf is claude code doing

@sacsbrainz @jarredsumner @kalomaze What is the command sir?

@stewones @jarredsumner @kalomaze It does

@jarredsumner @kalomaze Jarred why CC still doesn’t have a command to retrieve old sessions? This would make it easier to share bug reports.

@kalomaze Can you paste a screenshot of the terminal? Will investigate

@_mkahmad dafuq good job sir

Yay! I was rewarded $15,000 bounty from Apple. Account takeover via App Store 👀 #bugbounty #cybersecurity #pentesting

@andreeliasdev Claude Code é pra agentic workflows bro. Progressive dev tem que ser feito em IDE. Idealmente vc teria que usar os dois.

Cancelei o Cursor para testar o Claude e estou usando a umas 2 semanas E honestamente? Achei ruim demais a experiência de uso. Pô, eu quero fuçar meus arquivos, adicionar assets, escrever código na mão, rollback de código etc. e pelo terminal isso é horrível! “Ah André, aí é só usar junto de um editor de código” Pois é, o Cursor! Mas antes de vir xingar, tem pontos positivos: 1. Modelos mais baratos que pela API/Cursor 2. É isso.

@gustavo_pch Kkkkkkk

Ahh, a taxa é pra empresa, não pro consumidor, ufa