About

I'm Dylan, a solo developer. One person, making tools I want to use and games I want to play. Before becoming an indie developer, I spent 5 years in audit.

The most valuable skill in that profession isn't being clever — it's sitting still, looking closely, holding out. Two things became clear over those years:

First, most companies don't die from "not smart enough" — they die from "building a pile of things nobody wanted".

Second, on the job I could often see — many heavy processes could level up with just a small software tool. But writing software used to be a wall I couldn't get over.

I could see the solutions but couldn't reach the implementation.

Last year I started using AI Agents heavily. I realized Agents were no longer a chat toy — they could turn the "tool I think should exist" into actual code, actual product. That wall in front of me got thin. Since then I stopped talking about "starting a company" and just started building. Six products live on this site today.

About the name

awkthole is two words I jammed together:

awk — an old Unix text-processing language. Plain, usually one-liners, but it does an enormous amount of dirty work. It survived not by being elegant, but by ruthlessly nailing "just enough".

thole — an obscure English word. As a verb: "to endure, to bear, to hold on". As a noun: the small peg on a rowboat that gives the oar a fulcrum. Without it, all your effort just swirls water.

Use the simplest tools to brace against your own awkwardness — and turn them into the fulcrum that pushes the thing out.

That's the residue of 5 years in audit: details decide what's real, endurance decides the outcome. I can't win by being clever, but I can win by sitting still and looking closely.

What Agents changed (and didn't)

Last year I talked to a lot of people about AI. Opinions split into two extremes — either "nothing changed" or "everything changed". Both are wrong. Four things I sorted out while building AwkThole:

① Agents made "idea → prototype" 100× faster, but didn't shave a second off "prototype → someone uses it". Getting a real user to open it, enter data, come back the next day — none of this has anything to do with AI. Same as thirty years ago.

② More software, but no more users. The average person still only opens seven or eight apps a day. Everyone's cranking out wheels — but very few are solving "which one should I actually use?"

③ I don't believe in the "one-person company" myth. Agents aren't colleagues — they're a more comfortable wrench. They accelerate execution, but they won't decide what to build, won't take the difficult customer call, and won't push you through the night you almost gave up. Use Agents as Tool, not Partner.

④ AI accelerates execution, so humans should pivot toward judgment and creation. One end: deciding what to build — the cheaper experiments get, the more expensive going in the wrong direction is. The other end: shaping a fuzzy pain point into something that actually runs. Agents don't know whether a shop owner has any energy left for their phone at closing time. This dirt-on-the-boots feel — only humans can call it.

Why now

2026-03-28: Zhang Xue Motors won the WSBK Portugal round, beating Ducati, Yamaha, and Honda by 3.685 seconds. First time a Chinese motorcycle maker hit the top of an international racing series.

That was the first time I'd heard of Zhang Xue. But what hit me wasn't the 3.685 seconds — it was a 20-year-old documentary clip: at 19, he ran a beat-up repair shop in a small town in Hunan and begged a local TV crew to film his stunts. They came, he botched it on slick mud. The crew packed up to leave. He ran after them:

"Don't try to talk me out of it. I'll chase your van — show up whenever you have time, and I'll perform."

They drove off. An hour later he appeared in their rearview mirror. Skinny, all-in, chasing a TV van through more than 50 km of rain.

3.685 seconds beating Ducati is the outcome. What actually happened is 20 years. That's what's easiest to forget in the AI era — Claude + Cursor can ship an MVP in 30 minutes, but anything that's truly going to stand still takes 20 years.

The tools changed. The ruler didn't.

100 people can ship a demo in 3 minutes. Only 1 person grinds an engine for 20 years. The world champions in any field are always the latter.

What I believe

People always pay for value. But value isn't "more features" or "fancy tech" or "I used the latest model". There are only two kinds of real value: time, or emotion.

So I draw three lines for myself —

• Don't ship for "I built it" — ship for "I solved a real problem"
• Don't chase "more features" — chase "it gets used"
• Don't pretend "one person is an army" — collaborate when it helps, hold out when it matters

About this site

"awkthole" is just the URL — the names I wanted were all taken. It's not a brand. The site itself is an Astro + Tailwind static site, hosted on Tencent Cloud Lightweight.

Contact

Email: hello@awkthole.com — bug reports, game feedback, project collabs all welcome.