The answer to ai-generated test costs is to run the model once and record the rest. tester-army's…
the answer to ai-generated test costs is to run the model once and record the rest. tester-army's e2e, which led github trending on oct 6, records agent-driven test steps and replays them with no model calls until the app changes.
keeping the llm out of the hot path is the whole pattern.
Context
TesterArmy's launch post of October 1, 2026 describes e2e as an open source TypeScript testing framework that mixes locator assertions with agent steps such as agent.act(). It says the model runs once per step: when an agent step passes with a recorded check, e2e caches its actions and replays them on later runs without calling the model, so a cached step costs no tokens.
The docs say each agent step has a deadline and a model-call budget, and that agent steps run on any AI SDK model or on a ChatGPT, GitHub Copilot or SuperGrok subscription with no markup on tokens. A write-up on aicoder.com dated October 5 says the repository gained 1,430 stars in a day to top GitHub trending, with 3,828 stars total, and that verified steps replay with no model calls until the app changes.
The record and replay design matches the launch post. The 'until the app changes' condition comes from the aicoder write-up and was not found in the launch post text read here, so how and when a cached step is invalidated is unsupported here, not refuted.
The note dates the trending lead to October 6. The aicoder write-up is dated October 5 and says 'today', and the launch was October 1, so the trending date in the note may be a day off. Which day the repository led is not confirmed from GitHub itself here.
The note's wording that the answer to test costs is to run the model once is the author's framing. Per the post, only steps that pass with a recorded check are replayed, and the first run and any changed step still call the model, so cost falls on repeat runs and not on the first.
Related work
- Introducing e2e: open source agentic testing for web, iOS, and Android (TesterArmy, October 1, 2026) ↗Primary source for the framework, the cache and replay behavior and the no-markup claim.
- How agent steps work (e2e docs) ↗Primary source for step budgets and deadlines.
- TesterArmy's open-source e2e tops GitHub daily trending (aicoder.com, October 5, 2026) ↗Secondary source for the trending rank, star count and the until-the-app-changes wording.
Watch next
- Read the e2e docs on cache invalidation. Compare the framework with other agent-driven test tools on cost per repeat run.
Sources
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 9 October 2026 at 01:02 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →