Privacy-first coding.

Our goal is to make code and the AI that writes it more efficient. Compare the same workload before and after each change.

TwolapInteractive demo
AI Chat

Simplify this function and check if it’s faster.

Twolap

Can I benchmark this function, simplify it, and run the same test again?

Original codeids.ts
export function uniqueIds(ids: string[]) {  const seen: string[] = [];  for (const id of ids) {    if (!seen.includes(id)) {      seen.push(id);    }  }  return seen;}
Original benchmark
PendingMedian time per call
  1. Batch 1Pending
  2. Batch 2Pending
  3. Batch 3Pending
  4. Batch 4Pending
  5. Batch 5Pending
Revised codeids.ts

Waiting for the baseline.

Revised benchmark
PendingMedian time per call
  1. Batch 1Pending
  2. Batch 2Pending
  3. Batch 3Pending
  4. Batch 4Pending
  5. Batch 5Pending
5,000 IDs · 5 measured batches per version

Public release coming later.

Measure each change

Run the same tests and benchmark before and after a refactor.

Self-host on your PC

Use a local model to keep your chats and code on your device.

Bring your own provider

Choose a hosted model with your own API key in the private desktop build.