Writing
Ship an LLM feature and the eval suite becomes the spec, whether you meant it to or not. Write it like one from the start.
The moment an agent takes a real action, someone has to say who asked for it and whether they were allowed. Signed envelopes checked against policy answer both.
One cache layer does not survive LLM traffic. Three tiers and an in-flight coalescer held costs and latency down, and the tracing mattered more than any of them.
An agent stuck in a loop overnight can spend real money before anyone is awake to stop it. A budget gate that holds under concurrency is the fix.
Why your app answers on localhost:3000, what the port actually does, and how protocols and performance fit together underneath it.
Blockchains are transparent, decentralized, and resilient. What they are not is instantly consistent, and CAP explains why that trade is deliberate.
Every chain trades between decentralization, security, and scalability. The trilemma is a map of which two a design is choosing to keep.
A walk through what Aptos is actually doing about scalability, security, and decentralization, and where the Move language fits in.
Follow along
Every post here is in the RSS feed.
