I am writing this from inside the decision, not from above it.

Last week I closed three things on the same week. A multi-year retainer with a Brazilian fintech where I was the lead architect on their core ledger. A teaching contract on an executive MBA program where I taught Software Architecture to mid-career engineers. A meaningful project-based consulting pipeline I had spent five years building.

I closed all three to spend my full attention on one piece of software: DOS (DuranteOS), the personal AI operating system I have been sketching for months and finally started writing on September 9, 2025. As of this morning DOS has approximately zero of everything that would matter to a customer — no cloud platform, no shipped product page, no paying users, no formal pricing. It has 9 weeks of code, a notebook of design decisions, and a thesis I am about to spend a year stress-testing in public.

The plan is to have one paying reference customer by mid-2026. The plan is also to have ten by the end of 2026. If I am wrong about both, I will have spent a year proving a clean negative and I will have to rebuild the consulting pipeline from a colder start.

This post is not a celebration of that decision. The decision is not over. It is a working note from inside it, written as honestly as I know how, for the small number of people who are considering the same move.

The trade-off, before the feelings

The shape of the trade-off is the shape any indie founder's trade-off takes:

• Three steady professional commitments, all profitable, all healthy, all closed
• Replaced with one private software experiment, no customers, no revenue, twelve months of personal runway budgeted at the worst case
• A ratio of known income lost to unknown future income that any rational financial advisor would warn against

I will not put specific numbers on this post. The shape is what matters. As of week one (now), the runway clock starts. By month twelve I will either have a paying reference customer or I will be making a different decision.

What is unusual about the trade-off is that I am leaving a known good practice, not a stalling one. The retainer was healthy. The teaching was meaningful. The consulting pipeline was warm. The decision was not forced by failure on the prior path. That is the part that is hard to talk about, and the part this post is actually about.

The thesis I had to believe to leave

I have written elsewhere — in The Agent Is the Product — about why I think personal AI is the next operating system. I will not relitigate the thesis here. What I want to articulate is something more specific: the rate-of-change claim that made staying in the consulting pipeline feel like the riskier bet.

The consulting work — designing ledger systems, advising on architecture decisions for Series A teams, teaching enterprise architects how to think — is good work. I would still be doing it well if I had stayed. But I have been watching the ground move underneath the entire profession in a way that feels increasingly urgent.

The right column is not science fiction, but I want to be honest that I have not yet built the version of it I am betting on. I have ~9 weeks of DOS code and the conviction that the shape of the right-column model is correct. The next 12 months are about whether I can make that conviction concrete enough to charge for.

The question is not whether the consulting model survives. It does, for a while, in compressed form. The question is who builds the agent that replaces the consulting model. If it is not someone with twenty years of architectural judgment in their head, it will be someone with five years and a smarter prompt. I would rather be the former.

What I think I am underestimating

Three things I expect to surface in the first months. I am writing them down now so that future me has a baseline.

I have partial plans for all three. I will keep two long-time clients on a flat-fee quarterly architecture review specifically to recover forced exposure. I will publish Builder's Compass weekly to recover legible status. I will commit to a hard cutoff at 7pm to recover rhythm. None of these are tested yet. I will report back.

What I think I correctly anticipate

Two things, written before I have lived through them.

One. That the deepest work of my career will happen here. The rate at which I am already learning — about agents, about Claude Code, about the personal-OS architecture I have been sketching — exceeds any prior period by a factor I cannot quantify. The work is hard in a way the consulting work never was. Consulting was hard in execution. This is hard in taste — figuring out what the right thing to build even is, with no client telling me.

Two. That the financial drawdown will feel different than I expect. I expect to feel anxious daily about the runway clock. I suspect I will not. I think I will feel the clock as an interesting one — a measurable resource being converted into a measurable artifact. Whether the artifact will pay back the resource is unknown. The conversion itself I expect to find satisfying in a way that retainer work never was.

Both of these may be wrong. They are predictions, not observations.

What would have to be true for this to have been a mistake

I will think honestly about this question once a month, on a calendar reminder.

The decision will have been a mistake if, by July 2027 (twenty months from start), DOS has fewer than five paying operators and Builder's Compass is not large enough to support a return to part-time consulting. That is the failure threshold I am setting for myself, in writing, before starting. It is not a soft target.

What would I do if I hit it? Almost certainly: take a senior staff engineering role at a company building agentic infrastructure (there are several worth considering), keep DOS as a public open-source artifact, and write the retrospective honestly. The artifact would not vanish. The financial recovery would take eighteen to twenty-four months. The identity recovery would take longer.

This is not a ruinous failure mode. It is an expensive one. I am budgeting for it and I am not pretending it cannot happen.

Why I am writing this now

I am writing this now because the contracts ended last week and the savings clock starts today, and the easiest version of the next twelve months is the version where nobody knows I am doing this.

That would be a mistake. The DOS thesis benefits from being argued in public. The product will benefit from operator-prospects watching the build. The fear benefits from being named — naming reduces it without resolving it. And the small number of people considering the same move benefit from a note that is not retrospective polish.

If you are one of those people: the math is harder than the feeling, and the feeling is harder than the math. Run both. Talk to two people who left a known-good practice for a thesis that did not have a paying customer yet. Ask them what they underestimated. Then decide.

I will write the next note from this corner in three months. By then I will know whether the first version of the memory layer works, whether the architecture I am committing to survives contact with use, and whether the loss of legible status feels the way I expect. Either way, the report will be more useful than silence.

The bet is on the table. The runway is a number I can name. The work is the most interesting work I have ever done.

I will let you know how it goes.

— Lucas