Max Pastukhov — developer of the Agent Factory
I automate small companies: I build and train agents and put them to work where you actually work — on your own server, in your own data, in your own working order.
I have been living off my own business software since 2003. I have been working with agents since 2025 — that is when the technology and the prices finally reached the level needed for practical, everyday use.
I really am one person, but “one-person company” understates it: the number of agents and what they get done keeps growing. Every day they handle the volume of work that used to take me from six months to three years.
This site, everything published on it and all of its channels run under the control of those same agents. Even the agents themselves are developed by the agents 🙂
My clients are owners of small companies with no technical team of their own, and independent specialists and experts who need modern tools and an extra pair of hands — enough to get a week’s routine done in an hour. Their day goes to the flow of requests, reports and documents; everything rests on them, and one more employee on the payroll is expensive.
Exactly the same problems have always existed in my own one-person business. That is where the Agent Factory came from.
A short history
- 1999 — first job, as a webmaster: I learned there was money on the internet.
- 2001 — that this money was there not only for the chosen few, but for software developers too.
- 2003 — that I could successfully sell my own programs to the whole world myself. My own online products have fed me ever since, with no other source of income.
- 2005 — that this was becoming much bigger than a one-person business.
- 2025 — that a human loses even to the weakest open-source models wherever the work is virtual.
I used to build internet-marketing products, including some of the largest keyword databases in the Russian-language web, then spent several years building financial software for the international market. I came to agents through machine-learning models in finance: from them to language models, and from language models to agents. That last step turned my work upside down: today I no longer build tools for people — I build agent employees and the tools they work with.
My biggest mistake
I watched plenty of niches collapse and still failed to imagine that software development and the software business would go through the same thing — literally within months. For a developer who loves the craft, it was unpleasant to feel it happening to me. The choice was between fighting windmills and joining the shift — I joined.
My second failure is simpler and more painful: missed deadlines and overestimating my own capacity. I have always been chronically optimistic about my productivity. Both mistakes are now handled the same way — by measuring instead of guessing. It no longer matters much anyway: agents can do a year of my work in a day, and where I used to keep lists of ideas and plans, those lists now turn straight into execution.
What I learned about ideas and markets
The main lesson: do not trust your own ideas and plans. Market research and analytics say more about a market than decades of experience. So work starts not with an idea or a plan but with research — what I think about a market does not matter; what matters is reality in numbers, facts and metrics. I hand the analysis, the decisions and the planning to the agents as well: they have vastly more context and better judgment than any living person.
Lessons from practice
- I no longer count writing code as my expertise: with today’s language models it stopped being one — any current model out-codes me. I picked a direction where my skills reach a bit further than code: process automation, marketing and sales.
- Data and how it is presented are what clients underestimate most. Without data an agent works by guesswork.
- What matters is data and skills, not models or technology. Garbage in, garbage out: with quality data about your business, even the weakest model makes smart decisions.
- A well-built agent closes most tasks out of the box — but the last stretch always takes control and retraining. That is where the line between fantasy and reality runs.
- Volume and cost of work beat polish: it is far more effective to throw tokens at tasks than to postpone them until the perfect moment or the perfect plan.
- Backups and isolation beat micromanagement: it is better to let an agent make mistakes and roll back than to spend your own time watching its every step.
What I do NOT do
- I do not sell courses or “how to make money on AI”.
- No artificial vendor lock-in: from day one you can download backups of everything — agents, scripts, skills and data. The factory is simply more convenient and more stable in practice than something you would assemble by hand at home.
- I do not sell a solution in search of a problem: you get exactly the same tools I use myself in my own business. If something does not work in practice, it either gets fixed or gets removed without mercy.
- I do not take custom development orders. I provide the tool and teach you how to use it. If you want it to “just work by itself”, you will be disappointed.
My philosophy
Since 2025 my focus has shifted: not tools for people, but replacing virtual work with agents outright — and building tools for the agents themselves. I believe agents can already take over almost any work done at a computer; what holds this back is not the technology but the inertia of large systems and their hierarchies. Smaller companies have none of that inertia, so they adopt agents faster — and I expect them to outrun the large ones, not the other way around.
How to work with me
Like with a living person: write on Telegram — I answer there almost always, without delays and on weekends. Email gets an answer less often; voice calls and voice messages, never — there is physically no time for them. When you have your own crowd of agents, you will understand 🙂
Where to read me
I sometimes share what I learn in my blog and on Telegram — I should do it more often. The only thing that holds me back is the volume of work involved in managing a crowd of agents.
Max Pastukhov, developer of the Agent Factory. This site, its content and its channels run under the same agents. Contact me — online almost around the clock.