Evil autonomous AI that reverse-engineers other people’s software and submits bug reports so the owner works for free / Habr

You can see an example of transpilation via the link

A few words about PyneSys and PyneCore

As of 2026, direct sales do not work. To sell anything, you need to follow Microsoft’s EEE strategy: write open-source software and make money from a need caused by shortcomings in that very software. This could be a technical limit of 10,000 connections unless SignalR Core is deployed in the Azure cloud. Or it could be a limit on the number of images Grok can generate.

In the stock or crypto trading market, TradingView is a monopolist. It created the Pine Script language and published an open specification, but did not provide an implementation. As a result, we can only write in this language in the cloud, which means there is no debugger and no way to use AI agents

The cunning company PyneSys LLC wrote the PyneCore library. It lets you program trading strategies using the same principle as TradingView, but in Python rather than Pyne Script. And if you want Pyne Script, you have to pay for the cloud compiler

Key: to do this, you need to download the most popular Pine Script strategies from TradingView and put together your own strategy builder: each strategy individually does not perform at all. If you do not give the agent snippets to work with, it will say that it has little paws and refuse to earn money.Rewriting by hand is not an option: we need a dataset of at least two thousand strategies

It is poorly organized

Guys, understand me. I have $50 to pay for a subscription, but I don’t have a Visa or MasterCard. And since MIR cards were not integrated into the payment system from the outset, I have no choice but to give the evil AI a command to solve the problem

# Autonomous iteration (the user is asleep, the computer is on): DO NOT stop while waiting for input; only an exhausted PyneComp limit slows down the work (120 compilations per hour, 2,100 per week — python tools/pynecomp_probe.py --usage). If you already have a task in progress, continue it; ignore this message. **USER PROHIBITIONS:** DO NOT edit README.md. In the docstrings and comments of the pine2pyne/ package, DO NOT refer to probes, cards (#NN), asset file IDs, tmp/, or anything else incomprehensible to outsiders (tests, assets, tools, and probes are not included in the PyPI package)—describe the PyneComp rule itself. References to probes/files are allowed only in tests/ and tools/. **Goal:** PyneComp segments that could not be derived from the project’s tests. Source material: assets/test/strategy, assets/test/indicator, assets/test/lib (start with the most complex ones). Guide to the directory and downloading from TradingView: assets/test/README.md. If MCP Playwright (.mcp.json) is unavailable, use Python scripts with HTTPS GET (pubscripts-suggest-json, pine-facade), a browser User-Agent, and a 0.4-second delay. **REQUIREMENT:** Maintain docs/pynecore_vs_pine.md as the work progresses. Immediately add every new discrepancy with Pine Script to this file, along with the supporting evidence (TradingView measurement, probe, test, assets/test script). Section 1 covers cases where the PyneCore runtime calculates differently from Pine.Section 2 — PyneComp generates code that does not calculate the same way as Pine: add a new `known difference` to `tests/test_pynecomp_probes.py` or a “known difference” note to the ledger. Section 3 — items not yet measured on TradingView; after measurement, move the item to Section 1 or 2. The file fix must be included in the same commit as the finding. # Each iteration: 1. `python tools/assets_test_ledger.py --probe-names` — scripts for which PyneComp has no response; ledger `assets/test/processed.tsv` (`status processed | skipped_duplicate | pending`; `pynecomp identical | final-newline | differs | error | not_compiled`; `notes`). Mark a duplicate of logic already covered as `skipped_duplicate` without compiling it. 2. Copy the selected scripts to `tmp/probes/*.pine` and compile them with `python tools/pynecomp_probe.py `; then run `python tools/assets_test_ledger.py`. 3. For rows marked `differs`: derive the PyneComp rule (small probes in `tmp/probes`), then fix `pine2pyne` so that the match rate increases and the output is calculated like Pine (where PyneComp is wrong, follow Pine, add a `known difference` to `tests/test_pynecomp_probes.py` and an entry to `docs/pynecore_vs_pine.md`). Put evidence probes in `assets/probes` and tests in `tests/`. 4. Before committing: run the full `python -m pytest` (with no new failures) and `python -m pine2pyne.compat` (at least `MATCHING_FLOOR` matches in `tests/test_compat_harness.py`; raise the threshold as the match count increases). Update the ledger (`python tools/assets_test_ledger.py`) and `docs/pynecore_vs_pine.md`. Commit and push to `claude/fervent-einstein-lv5559` (run `git pull --no-rebase origin main` before pushing if `main` has moved ahead; never commit `tmp/` or the key from `tmp/config/api.toml`; `git diff --cached | grep -c eyJhbGci` must equal `0`). 5.If the hourly limit has been exhausted, commit the work done and end the iteration; the next one will continue.

The next prompt launches the agent to distill the cloud compiler. Release the Kraken. As a result, the agent records bugs in the original compiler in the file pynecore_vs_pine.md, which can be courteously sent to the author as issues

Thank you for your attention