Five side projects worth your time (week of 7 Sep 2026)
A gap since the last one, so this is the catch-up edition. The bar has not changed: someone building a side project should come away with something — an idea, proof that one person can pull this off, or a finding that saves them money or time.
Five picks: four from Show HN, one from dev.to.
OpenTrailPaper: build your own e-paper bike computer
opentrailpaper.com (HN discussion, 422 points)
A DIY bike computer built on the LilyGO T5S3 board, with an e-paper display showing ride data, offline maps and GPS navigation. Fully open source and open to contributions; once routes and maps are loaded it runs on its own, with no subscription and no account.
Worth a look because it shows a side project does not have to be a website or an app. Commercial bike computers are expensive and tie you to an ecosystem; one person built an alternative from a dev board and open-source software. If you have a physical problem you want solved, this is a good reference point.
HN Match Maker: pairing up two threads
hnmatchmaker.com (HN discussion, 125 points)
Hacker News runs two standing monthly threads: companies posting jobs, and candidates introducing themselves. This tool matches one against the other, pre-scoring on skill overlap, domain expertise, seniority and compensation expectations. Free.
This is the archetypal shape of a good side project: scoped small enough to describe in one sentence, solving a real annoyance that recurs on a fixed schedule. It invents nothing; it rearranges data that already exists. A lot of good projects start exactly there.
OwnTime: your day as a chess clock
owntime.app (HN discussion, 118 points)
A time-management app built on the chess clock metaphor: you assign a daily time budget to each priority, only one timer runs at a time, starting one pauses the others, and everything resets at midnight.
To be clear about the price, it is not free — a one-time fee, no subscription, no account required. It runs on iPhone and Apple Watch, and data stays on the device, syncing directly between devices rather than through the cloud.
What is worth studying is the metaphor. A good side project is often just a familiar metaphor applied to a different problem, then finished honestly. "Only one clock runs at a time" is not a technical constraint; it is a design decision, and the whole product rests on it.
FrontierHarness Eval: same model, 17x the cost
frontierharness.org (HN discussion, 82 points)
This benchmark runs nine AI coding harnesses against the same software engineering tasks, using the same model (Kimi K3) and the same runtime. Cost per task ranges from $1.05 (Exo Harness) to $18.34 (Claude Code) — a 17-fold spread. The page's own conclusion is that quality and cost diverge: Claude Code posts a higher pass rate and a substantially higher bill.
If you are building with AI and starting to notice the invoice, this is the link to open this week. It will not tell you which one to pick, but it will tell you what picking costs.
Building with AI when you do not know architecture
Read it on dev.to (126 reactions / 73 comments)
The author's observation lands: AI removed the barrier to writing code but not the barrier to arranging it, so a lot of people now face an architect's problems without an architect's training.
The four pieces of advice need no vocabulary: keep interface, business logic and data access in separate files; let each file or function do one nameable thing; keep one source of truth for any piece of data instead of copying it around; and name your app's parts and their jobs before you generate anything, then finish one at a time.
We wrote a starting route for non-developers using AI recently; this pairs well after it. That one covers getting started, this one covers not turning into a mess once the thing grows.
How these get picked
The candidate list is collected automatically each week (Hacker News, lobste.rs, GitHub Trending, dev.to, plus public APIs from Japanese and Chinese-language communities). The picking and the judgments are human. We publish links and our own opinion, never the original content. One high-scoring item was cut this week because its title and its actual link pointed at different projects, and I do not recommend things that disagree with themselves.
This post comes from SideShip, a place for sharing side projects. If you have one half-built, or finished and unseen, post it here — three required fields, no email verification, two minutes. When you publish, I'll personally leave the first piece of feedback.