Point it
at what
matters.
For twenty-five years I built software and led consulting teams. Now I build my own agents: the one that runs my training and sleep over Telegram, the memory layer behind my writing, the systems that hand me back my mornings. AI collapsed the cost of expertise. What's left is the question of what you point it at, and I write about that every week.Twenty-five years building software for other people. Now I build my own agents. AI collapsed the cost of expertise — what's left is what you point it at. I write about that every week.
Claude Code for Non-Coders. Thursdays: one consequential idea, free. Tuesdays: the hands-on build, paid.
We automate tasks that enhance human flourishing. We resist automation that encroaches on human judgment.
That is not a position against AI. It is a position against handing over the one part of the work that was always yours: the judgment that made you worth listening to in the first place.
A handful of things I'm convinced of. The rest of this page is me trying to prove them.
Value should accrue to the person, not the platform.
When the work gets faster, it hands time back. A deck that took an analyst three days now takes an afternoon, and those saved days come loose and sit on the table. The fight everyone mislabels as AI versus jobs is really an older one, over who picks that time up. By default it rolls to whoever owns the business, quietly, without anyone in the building calling it a fight. It does not have to. Build so the freed value lands on you and the people you serve, not on whoever rents you the model.When AI frees up days of work, someone picks that time up — by default, whoever owns the business. Build so it lands on you and the people you serve.
Stay the operator, not the assistant.
The real risk is not that AI replaces you. It is that it quietly demotes you. You feed the machine, it makes the call, and you sign off on whatever it hands back. The writer Cory Doctorow named that pattern the reverse-centaur, and I have watched capable people drift into it without noticing. Keep your hands on the decisions that carry weight, and pass off the rest on purpose. Sometimes the model really is better at a task than you are. The skill is being honest about which times those are.The risk isn't that AI replaces you — it's that it quietly demotes you to sign-off. Keep your hands on the decisions that carry weight.
Own the architecture.
Owning your setup does not mean running everything yourself. Use the frontier model when it is the best tool for the job; I do. What keeps you independent is understanding how your system fits together and being able to move. When a vendor changes its price or its terms, you swap the model underneath and keep working. The labs are far ahead, and building your own from scratch is slower and worse. So use them. Just do not build on a foundation you can never leave.Use the frontier models — I do. Just understand how your system fits together so you can always move, and never build on a foundation you can't leave.
The gate is effort, not permission.
No one is coming to authorize you. The barrier was never a credential or a title. It is whether you will put in the hours to learn, and that part is yours to control. That is the good news and the hard news at once. Money and access still tilt the field, and an afternoon of practice will not change that. But the part that is genuinely yours to take has never been this large.No one is coming to authorize you. The barrier was never a credential — it's the hours, and those are yours to control.
This is the first tool an individual can own outright.
The printing press, the factory, the mainframe, the cloud. An institution captured every earlier jump in productivity, because every one of them needed an institution's money to run. A capable model on hardware you already own is the first that does not. The frontier still belongs to a few labs, and it will for a while. That is fine. Good enough and yours beats best and rented more often than the frontier wants you to believe.Every earlier productivity jump needed an institution's money to run. A capable model on hardware you already own is the first that doesn't.
None of this is the point on its own. Owning the tools only matters for what you can finally aim them at: your health, your family, the problems that money and access used to put out of reach. Keeping your judgment valuable is the foundation. Building a life worth living is the point.Owning the tools is the foundation. Building a life worth living is the point.
The writing is the main thing.
Everything else runs downstream of it. I publish twice a week for senior technical people who built careers on judgment, stopped writing code years ago, and are working out what coding agents change for them. Thursdays are one idea developed in full, free and open to everyone. Tuesdays are the hands-on build, for paid subscribers.
Claude Code for Non-Coders
Twice a week. Thursdays are free and open to everyone: one consequential idea, developed in full. Tuesdays are the hands-on build, for paid subscribers.
Subscribe on SubstackD.E. Williams + Co. Insights
Where I first worked out the philosophy underneath all of this: using AI to protect your life instead of feeding it to your employer. The archive still holds up.
Read the archiveI don't just write about this. I run my life on it.
The argument only counts if I'm living it, so here is the stack I actually own.
LocalMemory
An open-source memory server for AI assistants. It gives Claude and other models memory that persists across conversations, running on your machine instead of a vendor's. Sixty people pay for it.
localmemory.coMy home agents
A Claude Code agent that runs my training, sleep, and health logging over Telegram. It's the clearest proof I have of the whole thesis: I built it, I own it, and it guards the mornings I care about instead of billing them to someone else.
Open source
zvec-go, agent-harness, othello-agent. The smaller pieces, built in the open, that the larger systems stand on.
github.com/danieleugenewilliamsWARE
A workforce-automation resilience assessment. It's a real product with paying customers, and it lives at automationresilience.com.
automationresilience.comAt scale, for others first
I did this at scale for other people before I did it for myself. That story is below, in About.
The writing comes first. This is what happens after you've read enough of it.
A few companies a year.
A few companies a year bring me in on their specific problem: what to build, what to buy, where AI actually belongs in the operation, and how to keep their people the operators instead of the assistants. I take the ones I can genuinely move. It's inbound only, and it sits downstream of everything I publish here.
In the room.
I also give talks and run workshops: keynotes for leadership teams, hands-on sessions for the people who have to build the thing. If you want me in the room, start here.
Twenty-five years building things for other people. Now I build them for myself, in the open.
I started as a software engineer on the platform that let the Federal Reserve, the World Bank, the IMF, and the BLS share statistical data. From there it was consulting. At Booz Allen I won more than $9M in new business and built a FISMA-compliant internal cloud that cut a research program's data-center cost by two thirds. Then Toyota and Lexus, whose digital properties I ran across a $35M portfolio. Then Pariveda, where I built a national practice from nothing to $6M a year and grew one account from $28K to $4M in nine months.
Somewhere in there the job changed. I stopped being the person who writes the code and became the person who decides what gets built and why. That seam is exactly what AI is now tearing open for everyone who made the same move.
So I do both again. I build my own tools, I write about how, and I advise the companies that want the same thing for their teams. I work in a window from early afternoon to evening, on purpose, because the mornings belong to my family, to fitness, to reading, and to being in the room instead of half inside a meeting in my head. I live on the coast of Virginia with my wife and two sons. Building a life worth living is the point.
- 2024 —D.E. Williams + Co.Builder · WriterOwn tools, in the open. Newsletter, agents, advisory.
- 2019–24Pariveda SolutionsManaging PrincipalBuilt a national practice $0 → $6M/yr; one account $28K → $4M in nine months.
- 2013–19BeyondsoftProduct LeadRan Toyota & Lexus digital properties across a $35M portfolio.
- 2010–13Booz Allen HamiltonAssociateWon $9M+ new business; FISMA cloud that cut data-center cost two-thirds.
- 2003–10Federal Reserve · SAICSoftware EngineerThe platform FRB, World Bank, IMF and BLS used to share statistical data.
Have something specific in mind?
One note for advisory or speaking. If I can genuinely move it, you'll hear back within two business days.
New here? Start with the newsletter.
One consequential idea, free every Thursday. Thirty-four thousand people read it.
Routes to Claude Code for Non-Coders on Substack. Unsubscribe anytime.