Barada's log

Voice latency benchmarks across OpenAI Realtime, LiveKit, Vapi, Retell, Pipecat, Cerebras.

Every time I correct my AI copilot, I'm leaking expertise - dark knowledge about how I think, what trade-offs I accept, where I draw the line.

How to build a TUI dashboard in Rust that monitors AI coding agents.

4.5 million fake GitHub stars.

AI coding agents don't choose libraries rationally.

Where does defensible value live in the models era?

Inspite of all the hype around personal assistants, personal agency still feels far from reality.

An open benchmark for agent infrastructure resilience.

Last week I let a swarm of AI agents run autonomously for a week.

We are the average of the friends we spend the most time with.

Predistribution changes how value is shared during AI-driven growth, before inequality hardens and redistribution becomes politically infeasible.

The productivity gains are flowing to capital holders instead of the workers generating them.

Inference workloads now account for 80% of AI compute spending.

When building becomes trivially easy, every software market becomes a red ocean.

Own the chips or the customers.

The build manifest: 90 days from foundation to production-ready autonomy.

Alignment as a runtime surface, policy enforcement without retraining.

Model outputs are hypotheses that need verification pipelines to catch errors before users do.

Memory and hands for the model: retrieval that doesn’t hallucinate.

The physics of production AI: latency engineering that keeps humans in the loop.

Before you build: the mental models for human-AI collaboration.

A 6-part series on building production AI systems.