Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →The roughly $20,000 figure is not the amount Uladimir Klyshevich says he paid. It is his estimate of what the tokens used to build his open-source agent harness would have cost at API list prices. He reports paying about $1,050 for subscriptions over three months. Those figures, and the project’s scope, come from his own account and have not been independently verified.
What does “40 billion tokens” mean here?
Klyshevich says he used 40 billion tokens over three months to build >_Fa, an open-source agent harness written in Dart. He describes the work as “40,000,000,000 tokens. 3 months of work. Zero lines of code written by a human.” That is his self-reported description, not an audited usage figure.
As an Amazon Associate I earn from qualifying purchases.
His stated token mix is approximately 24 billion cached-read tokens, 14 billion fresh-input tokens, and 2 billion output tokens. He says he used a mix of Kimi and GLM models, routing tasks between them. Based on that mix and the rates he cites, he estimates an API list-price equivalent of about $17,000–$20,000.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallCrashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWhy does he report spending only about $1,050?
The API-equivalent estimate and reported cash outlay answer different questions. Klyshevich says he paid about $1,050 for two Kimi subscriptions and two GLM subscriptions. He reports no cash cost for GitHub runners, releases, or Pages. He attributes the gap between subscription spending and the API-price estimate to subscription access, cached tokens, and sending tasks to lower-cost models.
The post does not provide receipts, usage exports, or a reproducible calculation. Treat both the approximately $1,050 payment total and the $17,000–$20,000 API estimate as figures he reports, not independently verified accounting. In particular, the API-equivalent number is a counterfactual price comparison; it is not a bill he says he received.
What did the harness reportedly include?
Klyshevich describes >_Fa as a Dart agent core with streaming providers, tools, sessions, compaction, and memory. He says it has command-line, desktop, mobile, and web interfaces, as well as browser-extension and CI uses. Other features he lists include on-device or browser-side execution, sandboxed scripting tools, git-backed memory, agent-to-agent messaging, session replay, runtime widgets, and declarative sandbox profiles.
He also says a separate factory project, dmtools-dart, builds and ships >_Fa releases, and that this release process has run without a human in the loop. These are claims about features and automation in his account; the post does not independently demonstrate their reliability, security, or the quality of the resulting software. Klyshevich also reports an approximately 7 MB CLI binary. His installation-size comparison is scoped to macOS arm64 and artifacts measured in September 2026, so it should not be read as a current, platform-independent benchmark.
So, is spending that much effort “crazy”?
Not necessarily—but the headline numbers cannot settle the question. The answer depends on what the work produced and what costs the comparison leaves out. If the builder values the resulting software, can use subscription access within its terms, and has a credible way to review and maintain agent-generated code, the project may be a worthwhile experiment. That is different from proving that the method is economical for another developer or a team.
- Cash paid: Klyshevich reports about $1,050 in subscriptions.
- API-price comparison: He estimates $17,000–$20,000 at list prices for the stated token mix.
- Value and quality: He describes a broad feature set and automated build-and-release workflows, but the post does not independently measure the software’s usefulness or quality.
The account does not quantify human supervision, failed work, maintenance, opportunity cost, or ongoing operating costs. Without those figures, the API-price comparison is not a return-on-investment calculation, and it does not establish that another project could reproduce the result for the same cash outlay.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What should other builders take from the example?
Use it as a first-person case study, not a budget forecast. Before comparing an agent harness—or deciding whether to build one—look beyond token totals and installation size. Relevant questions include how much human review the project needs, what security boundaries apply to tools and generated code, which providers and platforms it supports, how actively it is maintained, and what it costs to operate after the initial build.
Klyshevich’s account is striking because it combines a very large reported token total with much lower reported subscription spending and an unusually broad claimed scope. It is not evidence that tokens are free, that the software is production-ready, or that AI-generated development reliably replaces a team. The most defensible conclusion is narrower: this is one builder’s account of an ambitious experiment whose cash cost, API-equivalent estimate, and claimed results remain self-reported.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
