bruno

YEs, I do Cod3

Web2 Speed Runner Senior
Web3 50% in Progress to beat
AI Masterchief Proompt Engineer

[Renan Santos⬛️🟨⬜️ @RenanSantosMBL]

Existem dois caminhos

[bruno @brunos3d]

Kudos for the Clawdmeter developers/engineers
https://t.co/961QXcQehj

[bruno @brunos3d]

Saw the Clawdmeter project today and got inspired to build a Wear OS telemetry dashboard for Claude Code on my Galaxy Watch 🦞

Now I can monitor AI agent usage, reset timers, telemetry and a tiny pixel-art mascot directly from my wrist.⌚

Still prototyping.

[bruno @brunos3d]

I decided to turn all my research, experiments, and engineering discoveries into OptimizeDeals, an AI-native software engineering studio.

Lately I’ve also been using AI to write deeply technical and educational articles.

Latest article
https://t.co/kJQ6yT9Sru

[bruno @brunos3d]

Creating an ASS (subtitle file) generator with AI felt productive… until I had to explain it out loud. 🤐

[bruno @brunos3d]

Almost there

[bruno @brunos3d]

25/ Final thought:

Don’t outsource engineering judgment.

Use AI to accelerate execution and compress iteration cycles.

Keep ownership of architecture, standards, and decisions.

Build repositories for both HUMANS and agents.

That’s where leverage stays.

[bruno @brunos3d]

Most developers optimize for bigger context windows.

Useful, but the real leverage is usually architecture, task design, and token economics.

You don’t always need more AI.
You need better systems.

Here’s how I use Claude efficiently for real delivery. 🧵+

[bruno @brunos3d]

24/ Mental model:

AI throughput =
(Architecture × Task Clarity × Context Efficiency) / Noise

Noise is expensive.
Ambiguity is expensive.
Sprawl is expensive.

--> SUPER EXPENSIVE, especially when you use HIGH EFOOOORT