EvalAI
Cloud-CV's open platform for evaluating and comparing AI algorithms at scale — a decade-old Django codebase that used superpowers specs and plans to build its Yutori-driven benchmark scouting and outreach subsystem.
Reviewed last month · Jul 19, 2026
- Stars
- 2K
- Contributors
- 162
- Open issues
- 473
- Age
- 9.8 years
- Last push
- 3 days ago
- Latest release
- 1.1 Apr 13, 2018
View weekly commit data
| Week starting | Commits |
|---|---|
| Aug 23, 2026 | 0 |
| Aug 16, 2026 | 3 |
| Aug 9, 2026 | 8 |
| Aug 2, 2026 | 7 |
| Jul 26, 2026 | 21 |
| Jul 19, 2026 | 2 |
| Jul 12, 2026 | 0 |
| Jul 5, 2026 | 4 |
| Jun 28, 2026 | 12 |
| Jun 21, 2026 | 1 |
| Jun 14, 2026 | 8 |
| Jun 7, 2026 | 6 |
| May 31, 2026 | 13 |
| May 24, 2026 | 5 |
| May 17, 2026 | 9 |
| May 10, 2026 | 3 |
| May 3, 2026 | 1 |
| Apr 26, 2026 | 0 |
| Apr 19, 2026 | 9 |
| Apr 12, 2026 | 0 |
| Apr 5, 2026 | 0 |
| Mar 29, 2026 | 0 |
| Mar 22, 2026 | 0 |
| Mar 15, 2026 | 0 |
| Mar 8, 2026 | 0 |
| Mar 1, 2026 | 18 |
| Feb 22, 2026 | 17 |
| Feb 15, 2026 | 3 |
| Feb 8, 2026 | 10 |
| Feb 1, 2026 | 11 |
| Jan 25, 2026 | 15 |
| Jan 18, 2026 | 18 |
| Jan 11, 2026 | 16 |
| Jan 4, 2026 | 0 |
| Dec 28, 2025 | 0 |
| Dec 21, 2025 | 0 |
| Dec 14, 2025 | 0 |
| Dec 7, 2025 | 0 |
| Nov 30, 2025 | 0 |
| Nov 23, 2025 | 0 |
| Nov 16, 2025 | 0 |
| Nov 9, 2025 | 1 |
| Nov 2, 2025 | 8 |
| Oct 26, 2025 | 0 |
| Oct 19, 2025 | 0 |
| Oct 12, 2025 | 0 |
| Oct 5, 2025 | 0 |
| Sep 28, 2025 | 0 |
| Sep 21, 2025 | 0 |
| Sep 14, 2025 | 0 |
| Sep 7, 2025 | 0 |
| Aug 31, 2025 | 16 |
Metrics updated Aug 27, 2026 · collected automatically from the GitHub API
Spec structure
- Location
docs/superpowers/- Formats
- Markdown
- Notes
- Two subdirectories:
specs/holds two design documents (2026-06-06-yutori-scouting-design.md,2026-06-06-yutori-outreach-design.md) andplans/holds one large implementation plan (2026-06-06-scout.md) with checkbox task tracking and instructions for agentic workers to usesuperpowers:subagent-driven-developmentorsuperpowers:executing-plans.
Assessment drift: none
Spec-to-code drift
None observed — but the sample is a single change. Two design specs (Yutori scouting and outreach email pipelines)
plus one implementation plan landed in the same squash-merged PR #5119 as the complete apps/scout/ implementation
on 2026-06-14, so specs and code were born in sync. The only commit touching the app since is a regenerated
migration (PR #5130, 2026-06-30), which the plan already anticipated as makemigrations output. The repository
itself stayed active afterwards (about 26 commits on master between mid-June and early July), so the quiet spec
directory reflects a finished feature rather than an abandoned one.
Defects and rework
Not yet assessed. The specs record one design revision before implementation (both are marked “revised 2026-06-06
to move package to top-level apps/scout/ and replace JSON-file storage with DB models”), and the plan bakes
testing into each task (pytest, responses for HTTP mocking, jsonschema for schema validation), but there is
no post-merge defect history on the subsystem yet beyond the migration regeneration.
Maintenance outcomes
Too early to judge, though the adoption pattern is notable: a mature, multi-contributor platform (around 100
contributors, running since 2016) used superpowers to deliver an entire new subsystem — Django models, webhook
receiver, Celery outreach task, and management commands across 40 files — in a single reviewed PR. The plan names
the specs as its “source of truth” and targets agentic execution, making this a clean example of spec-first
feature delivery inside an established codebase rather than a greenfield demo.
Timeline
-
Follow-up on the spec-driven app
"Scout: Regenerate initial migration (#5130)" — the only commit touching `apps/scout/` since adoption; the spec directory itself is unchanged.
-
Superpowers adopted
Specs, plan, and the full `apps/scout/` implementation land together in one squash-merged commit — "Backend: Add scout app for Yutori-driven benchmark discovery and outreach (#5119)". This is also the latest commit on the spec path.