Fernando Maluf
@fernandocm199
Followers
712
Following
2K
Media
70
Statuses
431
Building useful stuff using AI
Sao Paulo, Brazil
Joined April 2014
I get DMs asking why I pay so much for Claude while I could use Codex or GLM (cheaper?). Truth: I spend nothing beyond the Max Plan. My LLM stack: $200 OpenAI Pro, $200 Claude Max, $200 Cursor Ultra, $1,000 API. No BS—here’s how to squeeze max value from CC. 0. Quick
2
2
29
Codex remains undefeated for long refactors. Step 0: make sure tests exist Step 1: plan Step 2: discuss Step 3: add checkpoints/checklists Step 4: let it cook Step 5: go nap, it’ll be done when you’re back.
0
0
0
Post @nextjs 16 problem: LLMs still trying to use export const dynamic / revalidate / fetchCache One agent even rewrote a revalidation hook thinking my code was “inconsistent.” Bro… you’re the inconsistent one 💀
0
0
0
Introducing the Vercel Marketplace for AI agents and services: • Unified billing + easy installation • Agents like CodeRabbit, Corridor, and Sourcery • AI services like Autonoma, Braintrust, Browser Use, Chatbase, Descope, Kernel, Kubiks, & MixedBread https://t.co/b0WNnNbYyw
vercel.com
Agents and Tools are available in the Vercel Marketplace, enabling AI-powered workflows in your projects with native integrations, unified billing, and built-in observability.
15
34
212
Upgrading most projects to @nextjs 16 this week. Cache Components are just 🔥🔥🔥 Opt-in caching at the function level, full PPR baked in — way better debugging and far fewer performance surprises in production. It just works.
0
0
0
Wild how fast “search the web” is turning into “work with the web.” Atlas feels like the first real agent shell.
0
0
5
GPT Browser is coming up!!!
0
0
3
Announcing our beta launch: X API pay-per-use model. We are expanding a closed beta to both new & power users who want to ship amazing apps on X. All selected users will receive a $500 voucher to build with the X API. 🤑💻🚀
323
284
2K
Super cool
Not gonna lie, this is pretty cool: Claude for Life Sciences: Anthropic launches Claude for Life Sciences, expanding Claude from a research assistant to an end-to-end scientific partner. The new Sonnet 4.5 model outperforms humans on key benchmarks (Protocol QA 0.83 vs 0.79) and
0
0
1
The AI SDK just keeps getting sharper — modular design, cleaner abstractions, and a dev experience that actually scales.
When I started working on the AI SDK in 2023 I thought it was a long shot to ever catch up with LangChain. Times change.
0
0
3
AWS US-East-1 down? It’s the IT equivalent of your coffee maker exploding mid-brew: Alexa stutters existential dread, Fortnite squads glitching into pixel purgatory, Venmo’s “payment sent” vanishing like a bad Tinder match. Lesson: single-region life is for amateurs.
3
0
4
The real story every time AWS goes down isn’t downtime — it’s how many systems quietly depend on it without knowing.
0
0
4
That 0.6T per tranche sets a high bar, but Tesla’s edge in autonomy and compute infra could make it feasible—capex efficiency and supply chain diversification will be the deciders.
To reach any of the 12 tranches in his proposed 2025 CEO stock compensation package, @elonmusk would need to grow $TSLA market cap by $0.6T… literally an order of magnitude more than any other Auto CEO has in 50 years.
0
0
4
Vercel just added zero-config support for NestJS backends. Now deploy full-stack: NextJS/NuxtJS frontend + NestJS backend in one go. Cuts setup from hours to minutes. For indie devs, this means faster launches without platform hopping.
We now support NestJS, NextJS, NuxtJS. Let me know what we should support next
0
0
5
About $ASML EUV machines: 100k parts from 5k suppliers, built over 25 years. China's mirror blanks? Great start, but the real moat is integrating it all without exploding your fab. Can't copy-paste decades of precision engineering.
1
1
4
Elon teased Grok 4.20 generalizing Python to other langs—prototype in Py, deploy in C++ without the rewrite grind. Early tests in Oct after Aug promise: xAI's small team prioritizing quality over rush. This could 5x dev speed for cross-lang projects. Worth watching.
0
0
1
People are underestimating how big this shift is. This diagram isn’t multi-agent orchestration — it’s a single Claude Code agent, with its entire state machine embedded inside one prompt. Each color represents a distinct internal state and decision path the model has to track
0
1
6
Elon's definition of AGI. Fair enough...
@kimmonismus Capable of doing anything a human with a computer can do, but not smarter than all humans and computers combined. That’s probably three to five years away. Grok 5 will be better at AI engineering (called “research” by the pompous) than @karpathy, for example.
0
0
6
Not running in background -> it works. when asked to to in the background, it crashes :(
0
0
2