#buildinpublic
Yesterday I came across Haptiker (https://t.co/rBxbgYolzT)
on Product Hunt. It lets you adjust your Mac’s volume by sliding along the trackpad edge.
Thought it would be fun to build my own, so I made EdgeVolume.
Still a work in progress, but yeah, here is the repo
https://t.co/0VBZCbdUVD
#opensource
Good morning everyone.
My morning started with
AI is making me look at Rust differently.
Not because it replaces Python, but because someone still has to build the infrastructure that makes AI fast, reliable, and scalable.
Python builds a lot of the AI.
Rust can help make the systems around it faster and safer.
I'm a MERN dev moving into AI, and I think it's time to stop building just another chatbot and start understanding what's happening under the hood.
Time to learn Rust.
Let's stay #connect along the journey
System Design Tip: Surviving the Cache Stampede
When caching at massive scale, TTL expiration can be dangerous. Without protection, a simple key expiration causes a massive spike in database load.
Two key strategies to remember:
Single-Flight / Mutex: Deduplicate concurrent in-flight queries.
XFetch: Asynchronously refresh high-traffic keys before they expire.
100%. If the migration isn't pinned to an immutable schema hash with a validated rollback receipt, you're rolling the dice in production.
We catch dialect drift by treating SQLite strictly as an ideation scratchpad, never the compiler:
The agent exports declarative HCL/schema code.
Atlas/pg-diff calculates the migration against the target engine's grammar, not SQLite's.
CI replays the generated SQL on a containerized target instance with production table statistics to estimate query lock times.
The reviewer gets the raw diff alongside an automated "Lock & Concurrency Impact" report.
Do you prefer forward-only migrations with feature flags, or strict up/down rollback scripts for automated agent changes?
If you give an AI coding agent direct access to a shared development database, you are begging for schema corruption.
Here is the architecture pattern top agentic teams use instead:
The Ephemeral SQLite-per-Session Sandbox
The Flow:
When an agent session initiates, spin up a lightweight, isolated SQLite / libSQL database in an ephemeral container.
Seed it with the production schema and synthetic seed data via an in-memory script.
Grant the agent full root/admin permissions to mutate, drop tables, or test migrations inside this sandbox.
Why this wins:
Zero network latency between agent execution and database reflection (sub-1ms local I/O).
If the agent hallucinates a catastrophic DROP TABLE or corrupts data, the blast radius is zero.
When the session completes, run deterministic schema diffing against the main repository branch.
If verified Generate clean migration file.
If broken Destroy the ephemeral SQLite container.
Never debug broken AI state on shared infrastructure. Isolate it at the process level.