Skip to content

Product and technology

Build, buy, or wait.
Argued by experts who disagree.

The database choice, the AI feature, the vendor quote, the rewrite nobody wants to own. She reads your specs in Notion or Confluence and your tickets in Linear or Jira, and a board argues the bet when it’s big.

Free to start. No card.

Example. You write: “Three questions: should our support AI answer customers directly or only draft for agents, vector search in our own database or a vector database, and is the observability renewal at $84k worth it?” She opens 3 rooms: AI replies, or drafts only? (The Data Scientist, up to 65 credits); Our database, or a vector database? (The Engineer, up to 65 credits); The $84k renewal (The CFO, up to 65 credits). Nothing runs until you tap.

Examples

The AI feature, the database,
the renewal.

The ask as an engineering lead would type it, how she read it, who sat, and what came back. Examples written for this page, not recorded runs.

Let the AI answer customers?

Work · Board of 3

We built an AI assistant for support. In testing it resolves 62% of tickets correctly, and 4% of its answers are confidently wrong. Do we let it reply to customers directly, or only draft for agents?

Chief of staff

Accuracy, trust and support load — worth a board. I’d seat The Data Scientist, The Customer and The Ethicist.

  • The Data ScientistAI/ML Specialist
  • The CustomerUser Voice
  • The EthicistValues Guardian

The board’s call

Drafts only for now; let it answer directly on the three ticket types where it’s right 95% of the time.

Four customers in a hundred getting a confident wrong answer is a trust problem. On password resets, invoices and shipping status, the error rate is a fraction of that.

Confidence73 · Moderate

The Data Scientist disagreed

“Your test set is 400 tickets. That 95% could move a lot on real traffic — measure for two weeks before you switch.”

First step Label two weeks of real tickets on the three typesOct 16

Example · Board of 3 · up to 135 credits

Our database, or a vector database?

Work · One expert

We need semantic search over 2 million support articles. Add pgvector to the database we already run, or adopt a dedicated vector database?

Chief of staff

The Engineer takes this one: the search you need now, and what each choice costs to keep running.

  • The EngineerTechnical Builder

Asked first

How fast does the article set grow?

SlowlyDoubles yearlyFaster

The call · The Engineer

pgvector in the database you already run; revisit at 20 million vectors.

Two million vectors fits in what you already run and back up. A second database is a second thing to page someone about.

Example · One expert · up to 65 credits

Renew at $84k?

Work · One expert

Our observability renewal came in at $84k, up from $61k. We use maybe a third of it. Renew, negotiate, or migrate?

Chief of staff

The CFO takes this — with what a migration would really cost your engineers.

  • The CFOFinancial Controller

The call · The CFO

Negotiate a one-year renewal at $65k or less, and price a migration for next year.

A migration in Q4 costs engineer weeks you’ve already planned. A competing quote in hand is also the strongest thing to bring to the renewal call.

Example · One expert · up to 65 credits

What she handles

Your specs, your tickets,
the design doc written.

She reads where the work is written down, asks what only your team knows, and writes the RFC when you ask.

Reads your specs and tickets.

Linear or Jira, Notion or Confluence, the Slack thread where it was decided. She says what she read; every write is a preview you confirm, on Pro and above.

Asked first.

When the answer depends on your system, the expert asks before answering. Being asked is never charged.

The RFC, written.

Ask for the design doc, the migration plan or the vendor email and it comes back as the whole piece.

One step per call.

Each room ends in one first step with a date. It waits under Open until it’s done.

The experts

Technical calls,
argued across models.

The Engineer on what it costs to keep running, The Security Chief on what an attacker could reach, The Ethicist on what your users would think if they read the decision. Board seats run on different AI models.

Pick an expert

Hover, tap or tab throughSwipe, then tap one
  • GPT
  • Claude
  • Gemini
  • Grok

Seats are spread across model families: GPT, Claude, Gemini, Grok and more.

Questions

Fair questions.
Straight answers.

What people ask before they bring her this kind of call.

Or see pricing · every use case

Does it understand our architecture?

As well as what you give it. Connect Notion, Confluence, Linear or Jira and she reads what a question needs, and says what she read. An expert asks first when the answer turns on something only your team knows.

Can it write the RFC or the design doc?

Yes. Ask for the piece and it comes back written — the RFC, the migration plan, the vendor email — ready for your team to edit.

Why different AI models?

So the argument isn’t one model agreeing with itself. Board seats are spread across model families; each researches alone, then reacts once to the others.

What if the experts are wrong?

They can be — they’re AI. That’s why an expert asks first when the answer turns on something only you know, why a board seats experts on opposite sides and on different AI models, and why every memo shows its confidence, who disagreed and what would change the call. The call stays yours.

What does it cost?

Free to start — 150 credits every month + 250 to start. Talking to her is free within a daily allowance. A consult’s card says up to 65 credits and a three-seat board’s says up to 135 credits; nothing runs until you tap, and a board never bills above its card.

Bring the technical bet before you make it.
Start free, with one message.

It waits in her box after you sign up. Nothing is sent until you press send.