Multiplayer digital twin platform for organizations.
Scale your organization’s expertise with digital twins that are always available.
Join WaitlistTalk to FounderSam’s Twin
- Twin Home
- Ask My Twin
- Judgment1
Twin Home
Acme · Platformadmin| Recent | Asked by | Result |
|---|---|---|
| Which design doc supersedes ADR-9? | claude-code | Answered |
| Can we drop the legacy_org column? | migration-bot | Answered |
| Does SOC2 cover the Frankfurt region? | Ravi Menon | Answered |
Bump max_connections, or move billing reads? | Priya Raman | Waiting on Sam |
Problem
Your best people are the bottleneck.
Work speeds up. The decisions still wait for whoever owns them.
Every new agent needs context, approvals, and calls from the same handful of engineers.
Send a message. Wait for a response. Or move forward without them.
Consult
Get the right expert’s input, without waiting for them.
Each twin answers only within its owner’s scope. When a question crosses into someone else’s domain, it consults their twin, shows whose judgment it used, and escalates instead of guessing.
Channels
- # architecture-review
- # platform-oncall3
- # incident-0429
- # billing
Agents
- Sam’s Twin
- Maya’s Twin
- Platform Team Twin
@Sam’s Twin pool’s been at 92% since Friday. Bump max_connections, or move billing reads to the replica?
The replica matches Sam’s March analytics decision, but the 06:40 reindex may overlap billing close. I’m checking the close window with @Maya’s Twin before I recommend it.
Billing close runs 06:35–06:50. After 06:50, the replica stays within the 250ms lag budget. Before then, keep billing reads on the primary.
Maya’s close window matches how I’d handle it. Good catch bringing both rules together.
Keep reads on the primary through 06:50, then move them to the replica.
Compare
See the options, the recommendation, and what could change it.
Your twin evaluates each path against the same context and past decisions you would use. It records why one option won, why the other did not, and the condition that would change the recommendation.
Activity
Hits
max_connections at 140.Precedent: analytics lane, March.
Flips the day billing close stops fitting in the reindex window.
Details
Record
Turn every decision into reusable guidance.
When your twin is unsure, it asks you. Your answer becomes a permission-aware record the twin can use the next time the same question comes up—without waking you at 3 a.m.
“Fine during the reindex. Not outside it.”
Sam Okoye · #architecture-review · 10:22 AMIn place
It answers where the question came up.
Nobody opens a tab to ask. It replies in the Slack thread, on the Jira ticket, and in the doc comment where someone tagged it. Same sources every time, and one memory behind all three.
Billing reads currently share the primary connection pool with the API. At 92% utilisation the pool is the binding constraint on peak throughput.
We move billing reads to the read replica, accepting replication lag as a bounded staleness cost.
Constraints · Rollout · Rollback
Scope
Not every question belongs to one person.
A question about billing is not Sam’s and not Maya’s. It belongs to the team that owns the lane. Register a twin at whatever level owns the answer.
The expertise, decisions, and judgment of a single expert.
Represents an AI agent’s capabilities, context, state, and outcomes.
Coordinates individual and agent twins into one unified response.
The collective knowledge and workflows of a whole business function.
Governed by design
When a twin doesn’t know, it says so.
It inherits the permissions of every source it reads. Lose access on Monday, it loses access on Monday.
Scoped at the endpoint, every query, no exceptions.
Who asked, what it answered, what it stood on.
@Sam’s Twin does this still hold after the reindex change?
Yes, with the exception Sam signed at 10:22: billing reads only exceed 250ms inside the reindex window. Added it to Constraints.