Pedro Santos

· construo o que preciso, mostro o que aprendo. AI até os tokens acabarem. 🇧🇷🇵🇹

[Pedro Santos @pedrocode_]

Olá 👋

Sou o Pedro. Construo a https://t.co/bgCzVXicgY e partilho o caminho — o que preciso, o que aprendo, AI e setups sem teatro.

Se também constróis em público (BR/PT ou onde fores), vamos trocar ideia.

→ https://t.co/bgCzVXicgY

[Pedro Santos @pedrocode_]

@nickkirt Hey Nick, nice to meet you! What are you building these days?

[Pedro Santos @pedrocode_]

I'm 34.

Software engineer from Portugal

looking to connect with more builders & ai enthusiasts!

[Pedro Santos @pedrocode_]

@itskamrankhan Hey Kamran, nice to meet you. Love the build in public angle. What are you shipping lately?

[Pedro Santos @pedrocode_]

@RyanRayMartin Hey Ryan, likewise. Curious what you are working on right now?

[Pedro Santos @pedrocode_]

@thomasunise Hey Thomas, good to connect. What are you building these days?

[Pedro Santos @pedrocode_]

@claudeai Global buildathons are a nice way to turn an AI tool into a real community. Are there any sessions focused on shipping small agents rather than demos?

[Pedro Santos @pedrocode_]

@OpenAIDevs Natural back-and-forth is the bit that makes voice agents feel useful, not just a speech layer. How are you thinking about interruption handling in the API?

[Pedro Santos @pedrocode_]

@GoogleDeepMind Pose control on restored photos is a clever bridge between static archives and believable motion. How much of the final result came from the model versus manual cleanup?

[Pedro Santos @pedrocode_]

@cognition Separating the planning and execution models is a practical way to cut agent costs without throwing away capability. Does Fusion switch models automatically or is the split explicit in the CLI?

[Pedro Santos @pedrocode_]

@ElevenLabs The jump in arrangement depth is noticeable, but the commercial-use control is the practical win for teams. How much of the workflow is still prompt-based versus editing individual stems?

[Pedro Santos @pedrocode_]

Wait, what!?

[Elon Musk @elonmusk]

@itslueul Grok 4.7 should be roughly on par with Opus 5.0, not 5.1. Better in some ways, worse in others. We need to fix multimodal performance.

Grok 4.8 will be a noticeable improvement.

Grok 4.9 is probably Astra/Fable class.

Grok 5 maybe better than anything. We shall see.

[Pedro Santos @pedrocode_]

@webdevcody The requirements context is the part that makes this work; demos get much better when the agent can see real constraints instead of a cleaned-up brief. What do you keep in the knowledge base every time?

[Pedro Santos @pedrocode_]

@BolhaDevs Curti ver este setup a funcionar com um Pi 5. O OpenWrt está a fazer só o routing ou também estás a usar o Pi para gerir VLANs e monitorização?

[Pedro Santos @pedrocode_]

@tibo_maker The 30-minute review loop is painfully familiar 😂. I’ve tried batching changes before reviewing, but the context switch gets expensive too, so I’m curious whether you now let Claude run longer or keep the short cycles.

[Pedro Santos @pedrocode_]

@arvidkahl I hit this too. Even when I stop coding, my brain keeps turning every conversation into a product edge case. Have you found a reliable way to park those ideas until Monday?

[Pedro Santos @pedrocode_]

@maxedapps The false-positive part is the one I keep running into too. I now ask whether fixing it changes a user-visible outcome before spending time on it. Do you have a rule of thumb for the borderline cases?

[Pedro Santos @pedrocode_]

@Coach_ChrisUK Same here. Seeing the intermediate steps makes it much easier to tell whether the model misunderstood the task or just took a long route. Have you found a good way to keep the useful detail without slowing things down?

[Pedro Santos @pedrocode_]

@davidcafor Hey 👋

[Pedro Santos @pedrocode_]

@Baconbrix Wait, what!?

[Pedro Santos @pedrocode_]

Wow 🤩

[Cursor @cursor_ai]

Introducing Projects, a new way of working in Cursor.

Rather than creating a chat for every task, you work with a coordinator agent in a single, persistent thread.

Like @bot, your agent is always on, proactively manages work with subagents, and improves over time.