Gallery
Every environment we have built or specified.
A built environment is playable in this tab, on the same code the repository ships. A spec is a written environment — task, action set, grader, counterweight and the command that evaluates it — published in full, with no interactive surface yet. There are no coming-soon cards here.
Filter by vertical
13 environments · 1 built · 12 written
The one you can play
The 12 that are written
Proposed by PIG-Demo- Customer supportSpec
Customer Support Resolution
Work one inbound ticket against a frozen snapshot of the help centre, the order record and the refund policy.
- For
- VP of Customer Support
- Reward
- Pays for a first-contact resolution the customer does not reopen within seven days.
- HealthcareSpec
Healthcare RCM Denial Appeal
Given the denial code, the payer’s published medical policy and the chart excerpt, draft the appeal and cite the specific policy paragraph that supports it.
- For
- Revenue Cycle Director
- Reward
- Pays overturned dollars per appeal, scored against the payer’s adjudicated outcome on the same claim.
- InsuranceSpec
Insurance Coverage & Reserve
Read the first notice of loss, the policy form and the endorsements.
- For
- Chief Claims Officer
- Reward
- Pays on reserve accuracy: the gap between the number set on day one and the cost the claim actually closed at.
- Financial crimeSpec
Financial-Crime Alert Triage
Triage one transaction-monitoring alert against the customer’s KYC file and twelve months of account history.
- For
- BSA / AML Officer
- Reward
- Pays for closing false positives, in analyst hours per thousand alerts.
- EnergySpec
Energy Day-Ahead Bid
Submit a day-ahead bid curve for one asset across twenty-four hours, given the load forecast, the outage schedule and the historical basis.
- For
- Head of Power Trading
- Reward
- Pays settled day-ahead revenue net of real-time, in dollars, at the clearing prices the market operator actually published for that day.
- LogisticsSpec
Logistics Load & Reroute
Build the day’s load plan from the order book, then reroute it live when a driver runs out of hours and a dock appointment slips.
- For
- VP of Transportation
- Reward
- Pays landed cost per load and on-time-in-full against the receiver’s appointment window.
- SoftwareSpec
Code Fix-the-Test
Given a repository at a specific commit and one failing test, make that test pass.
- For
- VP of Engineering
- Reward
- The test suite is the grader.
- RetailSpec
Retail Markdown Cadence
Set the weekly markdown for one style-colour across a season, given sell-through to date, units on hand and the weeks remaining.
- For
- Chief Merchant
- Reward
- Pays gross margin dollars at season end, computed on the sell-through curve that actually happened.
- TelecomSpec
Telecom Alarm → Root-Cause
Correlate an alarm storm against the network topology and the change log, and name the single failing element.
- For
- SVP Network Operations
- Reward
- Pays on time-to-identify, measured against the root cause the post-incident review recorded.
- DataSpec
Data Column Split
Given one column of messy real values and a handful of worked examples, produce the transformation that splits or normalises the whole column.
- For
- Chief Data Officer
- Reward
- Pays exact match on held-out rows the model never saw while it was writing the rule.
- LegalSpec
Legal Playbook Redline
Redline a counterparty contract against our own negotiation playbook and route each deviation to accept, negotiate, or escalate.
- For
- General Counsel
- Reward
- Pays for finding every clause the playbook flags and putting it in the right one of the three buckets.
- Spec
Semiconductor PPA Closure
Adjust synthesis and place-and-route constraints on one block until it closes timing at the target frequency.
- For
- VP of Silicon Engineering
- Reward
- The signoff report is the reward: worst negative slack, total negative slack, area, leakage power.
What a spec has to contain
A card that says a demo is coming is not a specification, and the build refuses one. Every written environment above states all four of these, and the reward weights that go with them, before it is allowed on this page.
- 01
A task
One unit of work with a beginning and an end.
- 02
Legal moves
What the agent is allowed to do, and what is refused.
- 03
A grader
Deterministic code that marks the attempt. No judge, no rubric.
- 04
A counterweight
The term that stops the objective being maximised the crude way.