PWV Portfolio Feed

Our $5M seed is led by PWV fund of Tom Preston-Werner (Co-founder of GitHub).

Joined by incredible investors and founders:
Luca (CFO, Apple)
Dane (CTO, Cloudflare)
Paul @kiwicopple (CEO, Supabase)
Amjad Masad (CEO, Replit) & Michele Catasta
Feross (CEO, Socket)
Zeno Rocha (CEO, Resend)

Your taste is transparent and stored in a readable taste[.]md file.

You can inspect it, edit it, or reset it entirely.

And you can share it:
$ npx taste push --all
$ npx taste pull ahmadawais/cli

Senior engineers can encode their patterns for everybody on the team.

The benchmarks speak for themselves:

CLI scaffolding: 4.2 edits → 0.4 edits (month 1)
API endpoints: 3.1 → 0.3
React components: 3.8 → 0.5
Test files: 2.9 → 0.2

The improvement compounds.

More usage = better constraints = fewer corrections.

"But what about rules files?"

Rules decay. You wrote them six months ago. Your codebase evolved 100 times since. Now they contradict each other.

Command learns continuously from what you actually do, not what you remembered to document.

Taste is auto-learned/managed/updated!

How? A new AI model architecture `taste-1`.

It's a meta neuro-symbolic model that separates learned constraints (symbolic) from generation (neural).

Your interactions become reward signals. The symbolic layer updates in real-time. No fine-tuning. No batch training.

I asked Claude Code and Command Code to write a simple CLI for today's date.

❌ Claude gave me console.log spaghetti.

🤘 Command Code knew I prefer TypeScript, Commander.js, semantic versioning, -v flag, and ISO date formatting.

Same prompt. Completely different output.

We call the missing piece "coding taste"

It's the thousands of micro-decisions you've built up over years: how you name variables, when you extract helpers, which patterns you reach for, how you structure tests.

LLMs can't see any of it.

They optimize for the statistical avg.

//Take command of your code

Models
Claude frontier LLMs
Opus, Sonnet, Haiku

Taste
`taste-1` meta neuro-symbolics
Learns how you code

Extend
CLI, Skills, MCP
Build on top of it

Code
10x faster
Ship features fast


Review
2x quicker
Slash review time in half


Bugs
5x slashed
Catch issues early

most documentation landing pages fail the squint test

if i can't tell what your tool does in 3 seconds with my eyes half-closed, you already lost me

documentation that looks like a product gets read
documentation that looks like an afterthought gets skipped

Appreciate this a ton. Maintainers shouldn’t have to bankroll the tools and infra that the whole ecosystem depends on.

Glad Netlify could help fuel your OSS adventures. And yes to this: if your company relies on open source, sponsor a maintainer you use and help gear them up. ✌️

back in Berlin for a few days 🥰

- live and one-shot weather dashboard (Open-Meteo)

Type safety baseline:
`strict: true`, `noUncheckedIndexedAccess: true`, zero `any`.

If you build TUIs in TS and want composable primitives, try it:

$ pnpm add terminui
$ npm i terminui

Build something and share with me. :)

So you can swap in a Node backend, a test backend, or custom transport without changing widget code.

Examples include:
- primary-screen dashboard
- alternate-screen multi-frame simulation
- kitchen-sink demo (all widgets + constraints + styles)

Stateful renderers exist where they should (selection/offset state), while rendering stays pure.

Backend contract is deliberately tiny:
`size`, `draw`, `flush`, `clear`, cursor get/set + hide/show

- additive/subtractive modifier composition via `patchStyle`

This makes style composition functional and predictable.

Built-in widgets:
Block, Paragraph, List, Table, Gauge, LineGauge, Tabs, Sparkline, BarChart, Scrollbar, Clear

- resolve non-fill constraints first
- distribute remaining space by fill weights

Then leftover cells go to the last fill chunk for deterministic results.

Style system:
- named ANSI colors + indexed(0..255) + 24-bit RGB
- modifier bitflags (bold, italic, underline, etc)