# On Solving Problems: A Working Philosophy
*H. William Welliver III, draft, April 2026*

---

I want to solve problems for the long term, sometimes really hard ones, in ways that actually solve them and don't create new ones along the way.

That's the whole thing (ed: maybe just "That's it"?), really. Everything else is detail.

---

## What this means in practice

**I start from the problem, not the solution.**

Every project begins with something that isn't working, or something that could work better. The temptation, especially right now with an industry awash in new tools and new enthusiasm, is to start from an answer and work backwards: pick the approach that's exciting, or fashionable, or fast, and then fit the problem to it.

I don't do that. It produces solutions that look right until they don't, and by then someone else is usually left holding the pieces.

**I respect what works, and I'm honest about what doesn't.**

Proven approaches became proven through hard experience: other people's failures, edge cases that didn't show up in the demo, years of production use. That knowledge is encoded in standards, established patterns, and the institutional memory of the people who built things before me. It shouldn't get discarded without a reason.

But the same discipline cuts the other way. Some approaches persist not because they work, but because they're familiar, or because questioning them feels risky. Identifying those, and having the courage to say so out loud, is part of the same job. Inertia has defenders, and "we've always done it this way" is a powerful social force. Pushing back on it is sometimes harder than resisting hype.

The goal isn't to protect the past or chase trends. It's to be honest about what's actually working, regardless of which direction the pressure is coming from.

**I use new tools when they earn it.**

I'm not a luddite. I've seen AI assistance genuinely speed up development work on certain classes of problems, in the right hands, with someone experienced enough to evaluate what comes out. I use it that way.

What I won't do is hand over judgment to a tool that doesn't have any. AI can generate plausible-looking answers faster than most people can review them, but it can't tell you whether the answer is right for *this* problem, in *this* domain, with *these* constraints. That's what experienced engineers are for.

The right question isn't "can AI do this?" It's "what's the right tool for this job?" Sometimes it's AI. Sometimes it's a proven library, a known pattern, a phone call to someone who's been there. Picking the right tool is itself a skill, one that gets better with experience and worse when replaced by ideology.

**I think about second-order effects.**

A solution that trades one problem for three smaller ones isn't a solution. Neither is one that works today and creates a maintenance burden that crushes the team in two years, or one that's fast to build but impossible to understand six months later.

Long-term thinking isn't pessimism. It's just accounting for the full cost.

**I'm comfortable with hard problems.**

Some problems look hard and aren't, once you understand them. Some look simple and turn out to be genuinely difficult. I've spent enough time in complex technical domains like networked audio, embedded systems, and industry standards to know the difference, and to be useful in both cases.

The hard problems are where I'd rather be. They're the ones that require real thought, where shortcuts reliably backfire, where the person who actually understands the domain is worth more than any amount of generated code.

**I let the work speak for itself.**

I don't think it's useful to broadcast capability. Good work speaks for itself, and the people who've seen it up close make better advocates than any amount of self-promotion. This isn't false modesty. It's confidence of a particular kind: the kind that doesn't need an audience to feel real.

It also keeps the ego out of the decisions, which matters more than it sounds. Pride of authorship is one of the most reliable ways to make bad technical choices. If you need a solution to be yours, you'll defend it past the point where you should have let it go. I'd rather help the right decisions get made than be credited.

If someone has a better idea, something that surpasses what I brought to the table, it would feel wrong to dismiss it for my own benefit. The goal is the best outcome for the project, not the best outcome for any one contributor's reputation. Being seen as an honest broker matters to me just as much as being seen as capable.

**I move freely because I'm not protecting turf.**

In complex collaborative environments, with clients, partners, standards bodies, and long-term relationships built on trust, someone needs to be the person who can say "their approach is better" without it becoming a political problem. That person can move freely because they're pursuing a result, not defending a position.

In environments where everyone is protecting their territory, the person who genuinely doesn't have turf to protect is a rare and useful thing. I've found that's often where the most honest conversations happen, and where the best work gets done.

---

## On the current moment

The world is moving fast right now, and a lot of the pressure on technical people is to move fast with it: adopt, generate, ship. There's a version of that which is genuinely good: removing friction, speeding up the parts of the work that don't require deep judgment, freeing bandwidth for things that do.

But there's another version that treats speed as the goal and everything else as an obstacle. That version produces systems nobody can maintain, codebases nobody understands, and technical debt that compounds until it becomes someone's crisis. Usually, that's not the person who made the call to go fast.

I've seen what happens when the people who know what good looks like get removed from the loop. It rarely ends well. The knowledge that lets you evaluate a solution, to look at it and just *know*, takes years to build and doesn't transfer easily. It's not in the documentation. It's in the judgment.

---

## What I'm not saying

I'm not saying slow is better than fast. I'm not saying new is worse than proven. I'm not saying AI has no place in serious engineering.

I'm saying: think it through. Ask whether this actually solves the problem, or just appears to. Ask what breaks in two years. Ask whether the people maintaining this will have what they need to understand it.

That kind of thinking isn't fear of commitment (ed: maybe "isn't a problem", instead?), it's the work.

---

*This is a working document, a view of how I think about my work, not a finished statement. The thinking evolves. The core conviction doesn't.*
