Felipe Martins

👨‍🎓 Analista de Sistemas
🎧 Vivemos esperando; Dias melhores; Dias de paz, dias a mais; Dias que não deixaremos para trás...

[Chase @0xChaseTM]

F*ck all $500 AI courses...

SpaceXAI engineer Lauren Tan showed how she automates coding agents

here’s how she builds verification into each agent run:

→ Keep the failing case with the patch

save the command, request payload or UI steps that expose the bug. retain the initial state, expected result and baseline commit

> use /create-verification-skill to capture startup and execution steps

replay the same scenario after the change

inspect actual outputs, correct remaining failures and rerun affected checks

→ Map the journey (including failure states)

for a form submission, record the route, permissions, test data and stable controls

valid input → expected submission and saved state
empty input → defined validation behavior
failed request → defined error and recovery behavior

use the feature’s actual requirements. trial the journey once, then update changed routes and controls with /maintain-verification-skill

→ Match each claim to its evidence

> Behavior - replay the real journey. retain outputs and screenshots.

> Performance - compare equivalent builds, inputs, workloads and cache conditions. retain profiles and measurement conditions

> Code quality - run types, lint and relevant regression tests. retain commands, exit codes and failure logs

bind results to the tested commit and worktree state. mark each check PASS, FAIL or NOT_RUN

→ Enforce feature boundaries in the import graph

Define public APIs and internal paths

For example:

allowed → features/cart
imports features/orders/index.ts

blocked → features/cart
imports features/orders/internal/*

resolve import aliases when checking dependencies

test both cases: the forbidden import must return a nonzero exit code, while the permitted import passes

run the check locally and in CI

keep related feature code together so the structure reflects the enforced boundary

→ Evaluate the skill on fixed tasks

compare current and revised instructions on matched tasks

hold the repo revision, fixtures and model settings constant

Score four criteria:

read code / reproduce / verify / report missing checks

inspect failures, revise the procedure and replay affected cases.

retain the skill revision and execution evidence

limit conclusions to the cases and criteria tested

→ Expand autonomy after repeatable proof

(local proof → repeatable runs → cloud work)

Before expanding execution, confirm:

> verification works in the target environment
> permitted files, tools and actions are defined
> stop conditions, escalation and review rules are explicit

return the tested revision, cohesive diff, commands, results and evidence paths

Keep unavailable checks marked NOT_RUN with concrete blockers

resolve required prerequisites before treating the task as verified

Save this to make every agent run easier to verify

[akari @akarixbt]

This is f*cking insane. SpaceXAI senior engineer (ex-Cursor) just showed how one person runs an entire company on GrokBot:

I run 50+ GrokBot agents like a company. A Chief of Staff runs my day, 3 PMs hand out the work, 40+ workers ship it overnight, and I just check the results over coffee. That team does the job of 20 engineers"

In this 1-hour workshop, SpaceXAI builds that GrokBot team from scratch, step by step

GrokBot → Chief of Staff → PMs → 40+ Workers → Autonomous Company

At SpaceXAI, 98% of engineers already run a team like this, and the ones who do are getting $750K+ offers. The rest are about to compete with them

Worth more than any $5,000 agentic engineering course

Bookmark it and watch today

Then read how to build your own team of GrokBot agents in the harness engineering guide below ↓

[lauren @poteto]

here's how i shipped 2,500 PRs last month to production

this was originally supposed to be for Cursor Compile in London. i couldn't make it since i was livestreaming for Grok @Bot Galaxy so i'm making it available for free here on X! watch it on 2x speed, i talk slowly

[Hux @huxlab]

今晚别刷Netflix了。

去看这个SpaceX AI工程师1小时的课程。

这是我见过最清晰、端到端讲透GrokBot到底该怎么真正用起来的课。

从单agent开始，一路到loop、graph、多智能体协作、verification和agent-friendly架构。

不管你是从没认真跑过agent，还是已经天天在用GrokBot，看完都会把很多“原来如此”的点一次打通。

真正核心就这几层：

1. 99%的人只用1个agent，没用loop和graph，只发挥了1%实力

2. 真正的新stack是Chief of Staff + PM + 20+ workers全自主团队

3. 怎么用高质量skills让agent像真工程师一样干活（pstack）

4. 怎么把代码和架构改成agent-friendly（Dune、contract、CI）

5. 怎么用verification让自己真正敢把活交给它们，而不是一直盯着

大多数人花好几个月才拼出的全景图，这节课直接给你画完。

把时间空出来，这可能是你这个月最值的一堂课。

[danielhe4rt.php @danielhe4rt]

Eu sempre usei muita coisa do @PinkaryProject pra estudo nesses anos pq eu sei que é o laboratório de testes do @enunomaduro mas agora com o https://t.co/Wh3Vbfn3W0 opensource, tá sendo uma aula toda vez que eu abro.

Simplicidade sempre vence e o Laravel é pro nisso.

[CompozyOS @CompozyOS]

Did you guys liked the new boot system? All details matter for us ❤️

Download CompozyOS ✨

[Google Antigravity @antigravity]

Today, we are launching a new Knowledge Work interface layout tuned for research, writing, and artifacts: same agent harness, plugins, and MCP tools, without the terminal and git clutter. Select your layout in the General tab of Settings.

Our goal with Antigravity is to help the world deliver value through products and research. Doing that always starts with building the logic, which is why software development is at the core of the product. But turning that work into real outcomes takes every role working together with a shared language and understanding.

[NestJS @nestframework]

Moving an Express app to Nest doesn't have to be a rewrite.

New docs chapter, two paths:
- Run Nest on top of your Express app and move one route at a time
- Switch in one release, with tests pinning the API

Plus a prompt for your coding agent.

https://t.co/lJ0wxqw3h4

[bakigul @bakigulai]

Google'ın AI agent'lar için hazırladığı açık kaynak skill reposu olduğunu bilmiyorsun

23.000'den fazla GitHub yıldızı var.
Gemini, Firebase, BigQuery, Google Cloud ve daha birçok teknoloji için hazır Agent Skills içeriyor.
Claude Code, Codex gibi araçlarla kullanabiliyorsunuz.
İleride işinize yarayabilecek, kaydedilip incelenecek kaynaklardan biri.

İncele
https://t.co/mqaA3DBAQ6

[Pablo Silva @pablosilva]

No iFood, nós construímos um agente criador de agentes, pra ajudar o pessoal a criar bons agentes no dia a dia.

Eu fiz a mesma coisa no @OpenBot_, criei um Bot Builder e pedi para ele mesmo criar as instruções e as skills para ele ser um best in class em criação de bots eficientes e eficazes.

Claro, revisei as instruções dele e as skills que ele criou e por enquanto, nos testes, ele melhorou muito meus bots e o mais legal, fez tudo sozinho.

Olhou bot a bot, ele mesmo fez as configurações e instalou as skills em todos os meus bots. Também criou cenários de testes pra testar as instruções e as skills.

Estou observando ainda, mas meus bots estão muito melhores, gastando menos agora.

Pra quem quiser o que eu criei tá aqui o link

https://t.co/p5PaBZpHZC

Ainda em WIP, estou evoluindo ele toda vez que meus bots fazem algo errado.

[beamnxw ./ @beamnxw]

Before you blame Opus 5.5, learn these 15 official Anthropic commands

15 essentials to get more out of Opus 5.5:

less busywork → parallel agents → faster iteration

steal the cheat sheet ↓

01 /goal → keep working toward a defined finish line
https://t.co/JznJdmit2l

02 /batch → split big changes across parallel agents
https://t.co/wxjwlPZZZS

03 /subtask → delegate with the current context
https://t.co/TmCoXiL4ot

04 /deep-research → get research with sources
https://t.co/5QFUiNJbTN

05 /design → create editable artboards
https://t.co/vh8wvUNGfc

06 /plan → review the plan before code changes
https://t.co/iFY3a4UxbU

07 /memory → manage project instructions
https://t.co/3KXIdFjKAQ

08 /compact → compress context around key decisions
https://t.co/gruQHVjONK

09 /branch → explore a separate conversation branch
https://t.co/vh8wvUNGfc

10 /rewind → restore a tracked checkpoint
https://t.co/9kPcEhWXNB

11 /mcp → manage connected tools
https://t.co/MkXDn7Rl95

12 /loop → repeat checks in a running session
https://t.co/kBIlEWP1j6

13 /schedule → automate recurring cloud tasks
https://t.co/lmGnxpr1aT

14 /simplify → clean up code with parallel reviewers
https://t.co/vh8wvUNGfc

15 /remote-control → continue the local session from your phone
https://t.co/GrvSOh1M3Y

start here: /plan → /goal → /simplify

save this, then read the full guide for Opus 5.5 Harness Engineering ↓

[sankit @sankitdev]

if claude keeps giving you generic code and ugly ui, stop writing longer prompts.

give it a better skill.

5 claude skills that blew up in october, and the actual problem each one solves:

1. matt pocock skills +35.3k
practical engineering guidance from matt pocock's own .agents directory. useful when "act like a senior engineer" isn't getting you good output.
https://t.co/VjWcrnwfBZ

2. ponytail +32.6k
makes your agent think like the laziest senior dev in the room. useful when claude overengineers a simple feature instead of writing less code.
https://t.co/OStZk6iCDH

3. security audit skill +31.3k
runs multi-phase security audits and returns independently verified, machine-readable findings. useful when "check my code for vulnerabilities" is too vague to trust.
https://t.co/3VYu4rwq0Z

4. impeccable +24.8k
gives your coding agent a real design vocabulary for polishing, reviewing and auditing ui. useful when your app works but still looks like generic ai slop.
https://t.co/Gwz3bYx7Yb

5. anatomy +23.1k
creates detailed interactive isometric illustrations of machines and complex systems, with svg, webgl and 3d effects. useful when you need to explain something visual but claude gives you another boring diagram.
https://t.co/QfTeojeLcM

bookmark this nerds, thank me later

[NestJS @nestframework]

Announcing the official AI module for NestJS 🧠

npm i @nestjs/ai

Register models by name, call them with timeouts, retries and fallback models, and let agents call your own provider methods through the guards and policies you already have.

https://t.co/s2bE2aL2lW

[CompozyOS @CompozyOS]

Need a Conductor replacement? Just use CompozyOS, it has much better features:

→ S-tier orchestration level
→ Always-on agents included
→ Built-in terminal
→ Loop engineering DAG system
→ Automations (jobs and triggers)
→ Multi-harness support in one place

and so much more...

[Google Antigravity @antigravity]

The Antigravity interface has a new look & feel. Many of you have already seen it, so let us know what you think!

[NestJS @nestframework]

📣 Announcing Production Rules in NestJS Observe: catch the bugs that never throw 🧷

A retry that charges a card twice, a query run once per row. Write down what a request may do - "checkout calls Stripe exactly once" - and every kept trace is checked against it.

[Laysa 💻🎲🌈 @LayTXT]

Postei esse projeto no LinkedIn e está me rendendo muitos novos contatos e DMs. Pra quem é muito novo de área de dados vou explicar em outras palavras o objetivo: Imagina que uma tabela começa a ter inconsistência em uma coluna, por exemplo um ID de 11 dígitos começa a receber 9 carácteres. Você passará a ter uma visão disso e como essa mudança afeta a qualidade de seus dados, além disso você também passa a ter uma levantamento por linguagem natural de possíveis causas para que sua investigação não precise começar do zero. Ganho de tempo e eficiência.

Agora vou testar o framework de uma outra forma, simulando um ecossistema de usuários. Em breve trago os resultados e como construi isso.

[Matt Pocock @mattpocockuk]

15 stars until we become the most-starred skills repo in the world...

...and the 13th most starred repo of all time.

https://t.co/UtVMpTxH2x

[NestJS @nestframework]

📣 Announcing release verdicts in NestJS Observe! 🚦

Every release now gets one: Regression, Changed, Clean or Not enough data. Under it, the evidence: new error groups, routes that went quiet, and the queries and outbound calls it added or dropped.

https://t.co/12Qj7Xaoe8

[Kamil Mysliwiec @kammysliwiec]

Announcing the official MCP module for @nestframework 🤖

npm i @nestjs/mcp

🧩 Tools, resources and prompts as provider methods, one decorator each
🔒 Your guards, interceptors and pipes on every call
🌐 Streamable HTTP from the app you already have

https://t.co/wxSgNLVAee