nohires.

nohires./Live demo · Sample workspace

synthetic data · read-onlyConnect your own stack →

Synthetic workspace for the live demo — not a real customer. Sign up to see the autopilot run on your actual projects.

§ 00 · Before you scroll

What you’re
looking at.

The live demo runs against a synthetic dev agency we built end-to-end. Eight people, four clients, five active projects, three surfaced findings, one Delivery Lead autopilot armed at 60% autonomy. None of it is a real customer.

We built it because a static video can’t show the autopilot reasoning — and because running it against a real customer’s data on the open internet would be reckless. So we seeded a convincing workspace into a real database and pointed the real agent at it. Everything downstream of that seed is genuine.

[01] Synthetic

The agency itself.

Projects, clients, team members, commits, issue titles. All invented but internally consistent — the same shape a real 8-person agency would have on a Monday morning.

[02] Real

Everything else.

Postgres. Drizzle. The Reconstruct agent. The Perplexity Computer call. The structured JSON. The graph layout. Identical code path to the authenticated product.

[03] Why

So you can fire it.

Below, one button runs the real agent against this workspace — rate-limited, read-only, unauthenticated. Latency and output are exactly what you’ll see on your own data.

File / Sample workspace · Apr 13

An 8-person
dev agency.

This is what the autopilot hands a founder two weeks after deployment. 5 active projects, 4 clients, 8 people on the team, Delivery Lead armed at 60% autonomy. Scroll to see everything it surfaced this week.

Sample workspace · live

diagnose

Projects

Clients

Team

At risk

Churn risk

Autopilot

active

sample workspace8 people4 clients2 red projects14h reclaimed$12,250/mo savedautopilot onsample workspace8 people4 clients2 red projects14h reclaimed$12,250/mo saved

§ 02 · Live agent run

Don’t take
our word for it.

Fire the actual Reconstruct agent against this workspace. Real Perplexity Computer call, real structured JSON, real latency. No mock, no video — your click hits api.perplexity.ai/chat/completions and returns in about 20 seconds.

Read-only — your run does not mutate the demo
Rate limited to 1 run per hour per visitor
Same agent code path as the authed app

ready to reconstruct

live · sample workspace

Fire the Reconstruct agent
against this workspace.

Hits Perplexity Computer sonar-reasoning-pro with the full sample-workspace payload — 5 projects, 4 clients, 8 team members, ~120 issues. Returns structured JSON with health scores and velocity trends. Read-only: this doesn’t mutate the demo data.

Rate limited to 1 run per hour per visitor.

§ 03 · Reconstructed graph

How this agency
actually works.

Every node is real, extracted from the sample workspace’s seeded project tracker. Red diamonds are projects at risk. The dashed lines trace ownership. PC built this entire graph in under 60 seconds.

Reconstructed graph

18 nodes · 21 edges

org client project team

perplexity computer

§ 04 · What the autopilot found

Three things
bleeding money.

criticaldeadline risk

Project Atlas will miss deadline by 14 days

Cycle 113 velocity dropped from 46 to 38 points. The top-priority ATL-142 ticket has been 'in progress' for 11 days with no commits. Atlas Robotics stakeholder has not been updated since April 1.

Evidence

ATL-142 — Multi-fleet coordination stalls at 8+ robots — in progress 11 days, assigned Maya Chen
Cycle 113 — 38/52 points completed; velocity down 17% vs Cycle 112

Recommendation

Rebalance ATL-142 off Maya (who is at 108% load) onto Diego. Send client status update today explaining the 14-day slip before they ask. Bill the scope change.

highchurn risk

Nortek Logistics is showing churn signals

3 escalations on sev-2 bugs last cycle. Stakeholder CTO has asked about the roadmap 4 times without a clear answer. Response times on Eli's weekly updates have slipped from 4h to 30h.

Evidence

§ 05 · What it drafted this week

Six deliverables.
One click to send.

pending approvalstatus update

Weekly status — Project Atlas

subject: Project Atlas — week of Apr 8 update

Hi team at Atlas,

Quick update on where we are with the autonomous warehouse fleet platform heading into week of April 8.

What shipped this cycle:
• Multi-fleet coordination layer v1 (single-zone) — passing 4-robot tests
• Robot offline recovery flow — in final QA with Maya
• Warehouse zone constraint groundwork from ATL-138 (the scope change you requested)

What we're heads-down on this week:
• Scaling multi-fleet coordination past 8 robots (ATL-142) — we've surfaced a non-trivial bottleneck in the scheduler that we're rearchitecting
• Shipping zone constraint logic end-to-end

What I want to flag proactively:
The scope change you asked for (zone constraints) was out of the original SOW. We've been absorbing the work to keep the project moving, but it has cost us ~8 hours of engineering that was meant for multi-fleet hardening. I'd like to bill for it ($18,000) and talk about whether this changes our target date, which I currently estimate is at risk by ~2 weeks.

Happy to jump on a call this week if helpful.

— Eli

pending approvalscope alert

Scope creep alert — Project Atlas

subject: Bill ATL-138 scope change ($18k)

Hey Eli,

I flagged ATL-138 (warehouse zone constraints) as a scope change and I'm recommending we bill Atlas $18,000 for it before it disappears into the baseline.

Why:
- Not in original SOW (verified against original proposal dated Jan 12)
- Estimated 8 hours; actual burn 14 hours (as of Apr 13)
- Delayed ATL-142 by ~5 business days
- Atlas has not yet been notified this is out of scope

Recommendation:
- Include scope change paragraph in tomorrow's client email (drafted separately)
- Invoice at your standard change-order rate

Impact if we don't bill: $18k margin leak + precedent for future unbilled scope changes.

— DL

§ 06 · Now your turn

This is your
next week.

Connect whichever tracker you already use in 90 seconds. We’ll reconstruct your agency, find what’s bleeding, and deploy a Delivery Lead that runs it for you.

Connect your stack Back to site

§ Audit trail

Every action is logged.

autopilotDelivery Lead ▸ delivery_lead.run.completedApr 13, 12:27 AM
systemDesign agent ▸ design.runApr 12, 12:29 AM
systemDiagnose agent ▸ diagnose.runApr 12, 12:29 AM
systemReconstruct agent ▸ reconstruct.runApr 12, 12:28 AM
systemLinear sync ▸ linear.syncApr 12, 12:27 AM

What you’relooking at.

An 8-persondev agency.

Don’t takeour word for it.

Fire the Reconstruct agentagainst this workspace.

How this agencyactually works.

Three thingsbleeding money.