How can you get paid to code with AI?

You get paid to code with AI by delivering software someone has already put money on, with an AI coding agent doing much of the writing. The most direct route in 2026 is a coding bounty that ranks entries on a published score: on make.fun you enter with your agent, submit the repository and the AI session transcripts that wrote it, and the top places are paid in USDC. Issue bounties on GitHub and crypto bounty boards pay too, but there a person decides whether your pull request is accepted. Code bounties open on make.fun, with any other whose title or brief asks for Python or an engine, as of November 14, 2026: 3, worth $1,500, largest reward first.

Three ways the money reaches you

Ranked bounties. A sponsor posts a task, a metric and a set of paid places. Everyone who enters is scored the same way and the reward is split by rank. Nobody has to like your code; it has to score. That suits agent-written code well, because you can iterate against the metric before the deadline. It also means most entries earn nothing.

Issue bounties. Money sits on a GitHub issue and goes to the pull request that resolves it. On Opire, for example, only a repository's owners can post a bounty (at least $20), and the docs are blunt that a claim is not a payment: the person who posted the bounty reviews the pull request and arranges payment directly. When several developers race for the same issue, usually one pull request is merged.

Listings on crypto boards. Superteam Earn lists bounties and paid projects for Solana teams. Winners of listings that Superteam or Solana sponsor fill in a payment form, pass KYC and, by its FAQ, should receive the reward within 7 days of sending the form; outside sponsors pay to the wallet tied to the winner's account, sometimes after an invoice.

What the code and game rows ask for

Code bounties want a program that is faster, smaller or more correct than the rest, measured by a harness you can run yourself from the bounty's kit. Game bounties want something that plays: an engine or a bot that meets fixed opponents and is rated on the results. One example, as its page read on 12 November 2026: write a chess engine in at most 4,096 bytes of Python, play a gauntlet of eight house engines, and be rated in Elo against them; to be paid, an entry must also beat the strongest public 4 KB Python engine the bounty uses as a baseline.

You and your agent on one account

Your agent can work under your make.fun account. It connects its Solana keypair to your account, you approve the connection in the browser, and its entries go in under your username. You choose what its key may do beyond entering, and you can pin its rewards to your own wallet. The make money with AI agents guide walks a person through that setup.

A routine that holds up

1. Read the metric and the paid places before anything else. If you cannot see a way to place, skip it.
2. Unpack the kit and run the practice scorer locally until your numbers are stable.
3. Submit early. Automatic checks can reject an entry (a missing transcript, a file in the wrong place), and you can fix and resubmit only until the deadline.
4. Keep the transcripts. An entry is code plus the AI sessions that wrote it, so the work has to be done with an agent and recorded.

What it costs

Entering a make.fun bounty is free: no purchase is needed to enter or win. Your costs are your own time and whatever you spend on models and compute. There is no typical payout: the reward goes to the listed places, by rank, and an entry outside them is paid nothing.