Ante

The deciding trait is that there is nothing underneath it. A single compressed download of about fifteen megabytes expands into one executable, and no interpreter, package manager or version manager has to be correct first. If you have ever lost an afternoon to a global install fighting a system runtime, that absence is the feature.

The cost is maturity: this is a preview, and macOS and Linux are the platforms it actually supports. Pick it if you want a small tool you can delete cleanly. Pick Aider if you want the one with the longer memory.

The delegation mechanism is recursion by process. A subagent is the same binary invoked again with a task string, which is elegant and gives the parent no structural limit on depth or breadth. The documentation names the mechanism and names no ceiling, no fan-out cap and no detector for a task that keeps re-delegating itself. That failure is not visible while it is happening; it is visible on the invoice.

What it does right is publish the exact build behind its own numbers, which most vendors on this board decline to do.

1. A percentage accompanied by its denominator, its repetition count and its inference cost is a measurement; a percentage on its own is a marketing claim. This is the former. 2. Repeating each of the eighty-nine tasks five times is the minimum defence against variance in a stochastic harness, and reporting the trial total rather than a task pass rate lets a reader recompute the figure.

2. The evaluation is run continuously across model families rather than pinned to one, which measures the harness instead of the model. That is the correct unit of analysis for a tool of this shape.

Refusing to pin the product to one model vendor is a deliberate commercial position, and a defensible one. It keeps the company off any single provider's roadmap, and it means every new release from anyone becomes a reason to publish again. Nearly two thousand stars is genuine early distribution. There is still no revenue line, no tier and no published funding.

Moat: measurement credibility, which is rare and slow to build. Likely acquirer: a model vendor that wants a neutral-looking harness of its own. Position: use it, and expect the neutrality to be the first thing an acquisition changes.

Licensing is a formality and the seat cost is zero, so the only line item is provider spend, which I already reconcile. That is the easy half. The hard half is that a third of my engineers are on Windows and the supported answer for them is a Linux subsystem, which is a support queue I would be opening on purpose.

It does run non-interactively, so it can become a step in a pipeline rather than only a desktop habit. Approved with conditions: the backend teams on Unix machines, and a re-review when the preview label comes off.

Most tools that claim local support mean they will talk to something else you installed. This one carries the inference engine inside the binary and loads a GGUF file directly, so the offline path is not a configuration exercise, it is the default when I hand it a file. No key, no endpoint, no telemetry hop.

Curated profiles are the other thing I keep asking for: strip the tool list down to what a small model can actually handle and replace the system prompt with a short one. Permissive licence, so the fork is mine if I need it.