The published database
What the data says
163 sourced stories
One story is one published database entry, not one unique team, project or successful shipment. The same organisation can appear more than once. Research and code-review systems are included alongside builds.
Calculated from the published data at build time. Latest source-check or story-update date: . This is not the publication date of every story.
Download the story data · Download these counts · How stories are chosen
Who published the source?
Each story has one source label. Builder means the person or team describing their own work, including a personal blog or paper. Vendor means published by a tool supplier or its owner. Independent means third-party research or reporting; it is not a guarantee of independent verification.
| Label | Stories | Share of all stories |
|---|---|---|
| Builder | 112 | 68.7% |
| Vendor | 39 | 23.9% |
| Independent | 12 | 7.4% |
Tools named in stories
Each tool is counted at most once per story. A story can name several tools, so tool counts and percentages can add up to more than the total. These are mentions in this collection, not market share or a ranking of quality. Unstated is the label for sources that do not name the coding tool used. Entries with only in-house tools and no catalogue tag are counted separately. Models and in-house systems are not added to catalogue tool counts.
| Label | Stories | Share of all stories |
|---|---|---|
| Claude Code | 78 | 47.9% |
| Unstated | 30 | 18.4% |
| OpenAI Codex | 28 | 17.2% |
| Cursor | 19 | 11.7% |
| GitHub Copilot | 12 | 7.4% |
| Claude | 9 | 5.5% |
| Devin | 6 | 3.7% |
| OpenCode | 6 | 3.7% |
| Amazon Q Developer | 3 | 1.8% |
| Claude Agent SDK | 3 | 1.8% |
| AICode | 1 | 0.6% |
| Amp | 1 | 0.6% |
| Factory Droid | 1 | 0.6% |
| Gemini CLI | 1 | 0.6% |
| Google Antigravity | 1 | 0.6% |
| Goose | 1 | 0.6% |
| Grok CLI | 1 | 0.6% |
| Jules | 1 | 0.6% |
| Lovable | 1 | 0.6% |
| Mistral Vibe | 1 | 0.6% |
| Niteshift | 1 | 0.6% |
| Rovo Dev | 1 | 0.6% |
| Sourcegraph Cody | 1 | 0.6% |
7 stories have only in-house tools recorded and no catalogue tool tag. They are not folded into Unstated. See the full data for those systems. Tool rows overlap; the other tables each sum to 163. Percentages are rounded to one decimal place.
Kinds of work
Each story has one editorial work-type label for its main focus. A project can involve other kinds of work too. Code-review stories use the existing labels; they are not a separate measured work type.
| Label | Stories | Share of all stories |
|---|---|---|
| New product | 45 | 27.6% |
| Company rollout | 28 | 17.2% |
| Migration | 25 | 15.3% |
| Language port | 14 | 8.6% |
| Refactor / maintenance | 14 | 8.6% |
| Upgrade / modernisation | 14 | 8.6% |
| Research / benchmark | 13 | 8.0% |
| Feature | 6 | 3.7% |
| Test migration | 4 | 2.5% |
Existing code or new code?
Brownfield means changing an existing codebase; greenfield means building new code; mixed includes both. Not applicable covers entries such as broad studies or rollouts where that distinction does not fit.
| Label | Stories | Share of all stories |
|---|---|---|
| brownfield | 104 | 63.8% |
| greenfield | 44 | 27.0% |
| Not applicable | 13 | 8.0% |
| mixed | 2 | 1.2% |
Techniques
148 of 163 stories map to at least one technique. A story can use more than one, so technique counts overlap and do not add up to 163. See the technique pages for the stories behind each count.
These are self-reported, curated success stories. The tags are what authors chose to mention. Counts show how common a technique is in this collection, not whether it works.
Context
93 stories use at least one technique in this stage.
| Technique | Stories | Share of all stories | What it means |
|---|---|---|---|
| Rules, skills and context files | 65 | 39.9% | The agent is steered with written instructions and context: AGENTS.md/CLAUDE.md, skill files, rule docs, migration guides, golden examples, indexed code or telemetry. |
| Spec or plan agreed first | 48 | 29.4% | A written spec, plan, design doc or brief is produced (often by the agent from old code) and settled or approved by people before implementation starts. |
Doing the work
94 stories use at least one technique in this stage.
| Technique | Stories | Share of all stories | What it means |
|---|---|---|---|
| Small, bounded batches | 37 | 22.7% | Work is cut into small PRs, components, phases or batches with capped size or volume, instead of one big-bang change. |
| Parallel agents on isolated copies | 32 | 19.6% | Several agents or sessions work at once, each in its own worktree, container, clone or branch, coordinated through git or tickets. |
| Orchestrator and worker agents | 27 | 16.6% | A lead, planner or orchestrator agent splits the work and hands pieces to worker or specialist agents, sometimes with a judge. |
| Deterministic tools plus LLM | 20 | 12.3% | Classic tooling (codemods, code search, static analysis, build graphs, scripted pipelines) does the mechanical or discovery part and the LLM handles bounded steps. |
| Background agents on in-house infrastructure | 26 | 16.0% | Agents run in cloud sandboxes or on internal platforms, triggered from tickets, Slack, CI or schedules rather than a developer's laptop. |
Verification
83 stories use at least one technique in this stage.
| Technique | Stories | Share of all stories | What it means |
|---|---|---|---|
| Automated verification | 83 | 50.9% | The agent's work is checked automatically: tests, compilers, linters, CI gates or fix loops, including parity harnesses, coverage or eval sets built before the change. |
Review and merge
69 stories use at least one technique in this stage.
| Technique | Stories | Share of all stories | What it means |
|---|---|---|---|
| Human review | 38 | 23.3% | A person reviews or approves the change (or the plan) before merge, or the owning team makes the final call. |
| Agent review | 36 | 22.1% | A separate agent or a different model reviews, audits or tries to break each change. |
| Risk-tiered routing | 9 | 5.5% | Rules that route each change by risk: low-risk changes auto-merge or get agent approval, risky areas go to a human. |
Release
7 stories use at least one technique in this stage.
| Technique | Stories | Share of all stories | What it means |
|---|---|---|---|
| Staged rollout of shipped code | 7 | 4.3% | The shipped code goes out gradually: feature flag, canary with rollback, pre-release, closed testing or a rehearsed data migration. |
Organisation
17 stories use at least one technique in this stage.
| Technique | Stories | Share of all stories | What it means |
|---|---|---|---|
| Enablement, spend controls and tool pilots | 17 | 10.4% | Company-level moves: preinstalled or self-hosted setups, peer and champion adoption, budgets and token policies, and pilots or proofs of concept of the tool before wider rollout. |