I hit the wall of rented
software. Then I built past it.
I'm Miles Huffman. I run a surplus goods company, I write the software that runs it, and I think out loud about where AI is taking operators like me. Those three things aren't separate careers — they're one continuous answer to a question I got stuck on years ago: why does the software I pay for control how my business works, instead of the other way around?
The operator
Eight years as COO and co-owner of Swiss Link, running operations, the P&L, and a punishing inventory problem — thousands of one-of-a-kind surplus SKUs across retail, wholesale, and a manufactured brand I built in the US.
The engineer
Self-taught, then production-hardened. I built the 30-module platform that now runs my company end to end — and I write the software for every engagement myself, not a contractor's bench.
The researcher
A techno-optimist who reads, writes, and thinks publicly about where agentic AI is taking operators like me — and why owning your stack matters more now than it ever has.
I didn't come to software through a computer science degree. I came to it through a problem I couldn't solve any other way.
For most of my career I've been an operator — COO and co-owner of Swiss Link, a Northern California institution that's been sourcing authentic military surplus since 1996. Surplus is the hardest inventory problem I know of: irregular lots, variable conditions, thousands of SKUs that never restock. On top of that I run a wholesale channel and Wavian USA, the NATO-issue fuel can brand I established in the States. Running all of that well is an operations and finance discipline before it's anything else.
Like every operator, I ran the business on rented software. A CRM here, an accounting tool there, a storefront platform, a dozen point solutions — each a reasonable decision in isolation. And like every operator, I eventually hit the wall: the data was fragmented across four vendors that didn't talk to each other, the costs stacked up quietly, and migrating off any one of them had become a project I kept deferring. I didn't fully control the thing my company ran on. That bothered me more than I expected it to.
So I learned to build. Slowly at first, then — when agentic AI coding arrived and the cost curve broke — fast. I consolidated a fragmented Salesforce, Accounting Seed, and Shopify stack into a single unified Odoo system the company owns outright, solo, in three months, under live operational load. Then I kept going: a 30-module FastAPI/React platform that now handles catalog publishing, inventory and margin reporting, and lifecycle email — built with CI, tests, and versioning, not throwaway scripts. No external SaaS manages those workflows anymore. I own the stack.
Somewhere in there, the practitioner turned into a thinker. The more I built, the more convinced I became that what happened to me isn't a one-off — it's the leading edge of a structural shift. Agentic AI is handing small and mid-sized operators the ability to own software they were locked out of for the entire SaaS era. I write about that publicly now, as an independent researcher and unapologetic techno-optimist, because I think most operators haven't yet updated their assumptions to match what just became possible.
Now I do this for other operators, part-time and on purpose. Not as a consultant reciting playbooks — as a practitioner solving the same class of problem for my own businesses every week. I keep a hard cap of three concurrent engagements so each one gets a real operator's attention. That's the whole practice: find the bottleneck, cut the rented complexity feeding it, and replace it with one system you actually own.
The software should serve the business — not the other way around.
Every tool you rent is a dependency you don't control. Dependencies are risk, and risk compounds.
Agentic AI broke the cost curve on building. The 'buy don't build' default deserves a serious re-examination.
Proof beats logos. I'd rather show you working systems than recite a client list.
I write more about all of this — the collapsing economics of bespoke software, where agentic AI is taking operators, and the case for technical sovereignty — on the blog.