Is Major Cooked? Free AI risk score + roast for your college major
Solo Era solo activities, solo maxxing, sidequests, challenge
Keepsake Project Management, Freelancers, DesignersSide Project Ideas for People With a Job: 12 Small Ones You Can Finish
The best side project ideas for someone with a job are small enough to finish in one to three months of evenings, solve a problem you have met yourself, and have one screen that does the whole job. Pick by time budget first: under 20 hours, or 20 to 60 hours. Anything that needs a community, a marketplace or a second person to be useful will stall.
What makes a side project idea finishable around a job?
Most side projects do not fail on skill. They fail on calendar. You have a few evenings and one weekend day, you lose one of them to a deadline at work, and a project that needed six uninterrupted weeks quietly becomes a repository you do not open. So the first filter is not “is this a good idea” but “does this survive a bad month”.
Four properties make an idea survive. First, one user and one job: the screen has a single input and a single output, so a half-finished version is already useful. Second, no cold start: it works with one user, today, without anyone else joining. Third, no data you must collect: you either already have the input or the user brings it. Fourth, a visible done line: you can say in one sentence what version one contains and what it does not.
Ideas that fail these four tests are not bad ideas, they are projects for a team. A dating app, a marketplace and a social network all need other people before they work. They are poor side projects and fine startups, which is a different decision with a different budget.
| Filter | Question to ask | Red flag |
|---|---|---|
| One user, one job | Can I describe the screen in one sentence? | “It depends on the user’s role” |
| No cold start | Is it useful with a single user and no network? | Needs a second side to show up |
| No data to collect | Does the user bring the input? | Needs a scraped or licensed dataset |
| Visible done line | Can I write version one on a sticky note? | The feature list keeps growing |
Which side project ideas fit in under 20 hours?
These fit a long weekend or three or four evenings, especially if an AI coding agent writes the first pass from a clear prompt. Treat the hour counts as rough and personal: they assume you have shipped a web app before and you are not learning a new framework at the same time.
01 Cron expression explainer
A single text box that turns a cron expression into plain English, and plain English back into a cron expression.
- Who it’s for
- Developers and ops people who touch cron twice a year and never remember the field order
- First version
- One input, one sentence out, five worked examples underneath
- Hardest part
- The reverse direction: ambiguous phrases like “every other Tuesday” have no exact cron form
- How to test demand
- Ship it, add a search-friendly page per common expression, and watch whether anyone arrives without being sent
This is the classic evenings project because the parsing is a solved problem and the value is in the wording. Decide up front what you refuse to translate and say so on the page. A tool that is clear about its limits earns more trust than one that guesses. See the idea on the board for the original pitch.
Read “Cron expression explainer” on the board and copy its build prompt →
02 Meeting cost ticker
A running counter that shows what a meeting costs in salary-hours while it is happening.
- Who it’s for
- Team leads and anyone who sits in recurring meetings
- First version
- Enter headcount and an average hourly rate, press start, watch the number climb
- Hardest part
- Not turning it into a joke app: the number needs a sober, shareable framing
- How to test demand
- Post it in one team channel and count how many people send it on without being asked
It needs no account and stores nothing, which keeps privacy questions off the table. The risk is novelty: people try it once. If you want a second use, add a share image for the final cost, not a dashboard. Read the original idea.
Read “Meeting cost ticker” on the board and copy its build prompt →
03 Was this actually on sale?
Paste a product link and see whether the “discount” is real against its own price history.
- Who it’s for
- Shoppers who distrust crossed-out prices
- First version
- One store, one product page format, manual price log you start recording today
- Hardest part
- History. You have none on day one, so the tool only becomes honest after weeks of logging
- How to test demand
- Offer a small waiting list for one store before you write the scraper
This one punishes impatience: the data is the product, and it accumulates on a clock you do not control. Start logging a few hundred products before you build any interface. It fails the “no data” filter, but only mildly, because you collect it yourself and quietly.
Read “Was this actually on sale?” on the board and copy its build prompt →
04 Subscription graveyard
A list of the subscriptions you have paid for and forgotten, with the yearly cost of each one.
- Who it’s for
- Anyone who found an old charge on a statement and felt slightly ill
- First version
- Paste or type your subscriptions by hand, get a yearly total and a cancel-this-first list
- Hardest part
- Bank connections are a legal and security project. Skip them; use manual entry or a pasted statement
- How to test demand
- Ask ten people to try it with their own list and see whether they come back after a month
Keep the first version offline: store the list in the browser, never on a server. That removes the hardest trust question in one decision. The original idea is on the board if you want to see how it was pitched.
Read “Subscription graveyard” on the board and copy its build prompt →
05 A group poll that closes itself
A poll with a deadline and a quorum, so it ends with a result instead of a thread nobody answers.
- Who it’s for
- Friend groups, clubs and families choosing a date or a restaurant
- First version
- A link, three options, a closing time, and a result page everyone can open
- Hardest part
- Reminders. A poll that closes itself still needs people to see it before it closes
- How to test demand
- Use it for your own next three group plans before showing anyone else
It breaks the “no cold start” filter slightly because voters must arrive, but the organiser is the one user and gets value alone. Anonymous voting keeps accounts out of scope. See it on the board.
Read “A group poll that closes itself” on the board and copy its build prompt →
06 Palette from a photo
Upload a photo, get a small palette whose text and background pairs already pass a contrast check.
- Who it’s for
- Designers and developers styling a page from a mood image
- First version
- Client-side colour extraction, five swatches, a pass or fail mark on each text pair
- Hardest part
- Taste. Extracted colours are often muddy, so you need a rule for nudging them without lying about the photo
- How to test demand
- Post three before-and-after examples and see which people ask to download
Everything runs in the browser, so hosting costs are close to zero and you never see anyone’s photos. The contrast rule is the part that makes it more than another palette generator. Original idea.
Read “Palette from a photo” on the board and copy its build prompt →
Which side project ideas take 20 to 60 hours?
These need a database, a small amount of real logic, and some care at the edges. Plan for a month or two of evenings. Use the MVP scope planner before you start, because this is the range where scope creep ends projects.
01 A changelog your users actually read
Turn commit history into a short, human changelog page that shows only what a customer would notice.
- Who it’s for
- Small product teams whose changelog is either empty or a dump of commit messages
- First version
- Connect one repository, group merged pull requests, let a human edit before publishing
- Hardest part
- Deciding what a user would notice. A model can draft it; a person has to approve it
- How to test demand
- Write one changelog by hand for a friendly team and see if they ask for the next one
The human-approves step is also what makes it finishable: you are building an editor with a good first draft, not an autonomous writer. The idea on the board is a good starting brief.
Read “A changelog your users actually read” on the board and copy its build prompt →
02 Kids chore board with pocket money
A shared board where children tick off chores and the pocket money total updates.
- Who it’s for
- Parents of children old enough to read and young enough to care about a number
- First version
- One family, a list of chores with values, tick-off by tap, a running balance
- Hardest part
- Children’s data. Keep it to first names and no accounts for the kids
- How to test demand
- Run it in your own house for a month; if the kids still tap it, show other parents
You are your own first user, which is the best testing setup a side project can have. See the idea.
Read “Kids chore board with pocket money” on the board and copy its build prompt →
03 Pickup game organiser
A page for amateur football games that tracks who is in, who dropped out and whether the game is on.
- Who it’s for
- The one person in every group who organises the game and chases replies
- First version
- A game link, in or out buttons, a minimum player count and an automatic “on” or “off”
- Hardest part
- Notifications that do not rely on an app install
- How to test demand
- Run your own weekly game through it for six weeks
The organiser is the only paying-attention user. Build for that person and let players tap a link without signing up. The original post.
Read “Pickup game organiser” on the board and copy its build prompt →
05 Move-in day documentation
A guided checklist for photographing a rental flat on move-in day, so the record exists when the deposit is argued over.
- Who it’s for
- Tenants moving into a rental
- First version
- Room-by-room prompts, timestamped photos, an exported PDF
- Hardest part
- Photo storage and the PDF layout; do not promise legal weight you cannot back
- How to test demand
- Offer it as a free checklist page first and see which rooms people skip
Rules about deposits differ by country, so keep the first version jurisdiction-free: it is a record-keeping tool, not legal advice. The idea on the board.
Read “Move-in day documentation” on the board and copy its build prompt →
06 Which env var changed between two deploys
Compare the environment variables of a working deploy and a broken one and list only what differs.
- Who it’s for
- Developers debugging a deploy that worked yesterday
- First version
- Paste two env files, see added, removed and changed keys, with values masked by default
- Hardest part
- Trust. People will not paste secrets into a web page, so run the diff in the browser and say so
- How to test demand
- Release it as a command-line tool first, where secrets never leave the machine
A tool that never sends data anywhere is both the safer product and the cheaper one to run. See the idea.
Read “Which env var changed between two deploys” on the board and copy its build prompt →
How was this list chosen?
This is editorial judgement against stated criteria, not data. Every entry passes the four filters above in some form, every entry has a single user type, and none needs a network of other users to be useful. Where an idea matches a post on the board, the entry links to it. Hours are estimates for someone who has shipped a web app before; they are not measurements.
Ideas that need regulatory approval, a licensed dataset, payments between strangers or moderation were left out on purpose. They are real businesses, but they are poor use of three evenings a week.
How do you pick one and check that it is worth building?
- Write the time budget you actually have, in hours, for the next eight weeks. Subtract a third for the weeks that go wrong.
- Choose the section that fits, then cross out every idea you would not use yourself. Using it yourself is the cheapest feedback loop you will get.
- Write version one on a sticky note: one input, one output, one sentence. If it does not fit, the idea is too big for now.
- Set a stop date. If the first version is not live by then, you drop it or cut scope; you do not extend the date.
- Show it to five people who have the problem. One who asks when the next version arrives is worth more than twenty who say it is nice.
For the detail on testing before you build, see how to validate an app idea. If you want an AI coding agent to write the first pass, the vibe coding project picker helps you choose a bounded project.
Project: [one sentence] One user: [who] One input: [what they provide] One output: [what they get] Time budget: [hours] over [weeks] Stop date: [date] Out of scope for version one: accounts, payments, notifications, settings, admin screens Done means: [one observable behaviour]
Where do you find more side project ideas?
Browse free app ideas on the board by category and keep only the ones you would use yourself. When you find one, the prompt on its page has the whole description in it; copy it, or send it straight to Claude Code, Codex or Cursor, and add the time-box brief above so the agent knows when to stop.
If you have an idea you will never build, post it. No account or email is needed, and it is permanent once published. For a weekend-sized list see weekend project ideas, and for projects built around an AI coding agent see Claude Code project ideas.
Frequently asked questions
What is a good side project for someone with a full-time job?
A good one has a single user, a single job and no dependence on other people showing up. It runs on your own data or the user’s input, and you can finish version one in 20 to 60 hours. Tools for a recurring annoyance you already have beat anything that needs a community.
How long should a side project take?
Aim for a first version in four to eight weeks of evenings, and set a stop date before you start. If it needs more than about 60 hours, split it and ship the smallest useful half. Projects without a stop date tend to drift until they quietly end.
Should a side project make money?
Not necessarily. Decide the goal before you build: learning, a portfolio piece, or income. Income needs a buyer who pays for a defined result, which changes what you build. If you are unsure, build it as a free tool first and see if anyone returns, then think about charging.
How do I stop abandoning side projects?
Shrink the first version until it fits in a weekend, ship something live in the first week, and tell one person when you will finish. Cut the settings, accounts and admin screens. Abandonment usually follows a stretch with nothing visible to show, so keep a working version online from the start.
Can I build a side project with an AI coding agent?
Yes. Write a precise prompt that names the user, the input, the output and what is out of scope, then review everything it produces. Agents are fast at the first pass and weaker at deciding what not to build, so the scope decision stays with you.
Created with AI assistance and edited for this site. Examples are illustrative and do not represent customer research. About the publisher.