Community guidelines

colift exists to produce real work with genuine public value. The guidelines below describe what that means in practice, what happens when a submission falls short, and why the stakes here are higher than usual.

Do real work

Every task asks you to do something concrete: visit a store and record prices, review a government website on a real device, transcribe a public document, document a sidewalk. You have to actually do it. Don't submit placeholder data, don't fabricate observations, don't fill in fields without completing the work.

The difference between "I went to the store but the item was out of stock" and "I made up a price" matters. The first is an honest finding. The second is fabrication.

Submit your own photos and evidence

Photos and attachments must come from your actual visit or work session. Don't use images from the internet, from prior visits by someone else, or from other tasks. The platform fingerprints incoming images against previously submitted ones (SHA-256 hash match) and flags duplicates. A detected duplicate goes straight to human review.

Don't fabricate findings

If you couldn't complete an item, the item had no price tag, the page you were supposed to review was down, the document was illegible, note that honestly. Almost every task spec anticipates this kind of outcome. "Not available" or "could not determine" is a valid finding. An invented number is not.

AI tools are allowed, with limits

You can use AI tools while working, to look something up, summarize a long document, or polish the wording of your notes. What you can't do is use AI to generate your entire submission without actually doing the task yourself.

The question the validator asks is: did you genuinely engage with the task? If the answer is yes and you used AI as a tool along the way, there's no problem. If the answer is no, if you asked an AI to fabricate the results and submitted them as your own, that's fabrication.

Time log honesty

The project timer measures your active working time. Don't leave a session running while you're not working. Sessions that don't match the work product, for example, several hours logged for a submission that could only take twenty minutes, get flagged for velocity anomaly and go to human review.

You're only credited for real active time, capped at a per-task maximum. If your measured time is below the cap, that's what you get. There's no gain from leaving the clock running.

One submission per task instance

If you made a mistake or want to update your work, use the "needs changes" flow when a reviewer asks you to, or contact us. Don't create duplicate submissions for the same task instance.

What happens when a submission is flagged

Flagged submissions go to a human reviewer at the sponsoring organization instead of auto-approving. The reviewer can:

Approve — hours are credited (the amount may be reduced if quality is partial).
Request changes — you can revise and resubmit.
Reject — rejected work earns zero hours. No partial credit.

Repeated violations can result in account suspension.

Why the stakes are higher than usual

Certified hours appear on Form CF 888 or equivalent documents, which are formal legal attestations. False attestation on federal benefit documents carries exposure under 7 U.S.C. §2024 and 18 U.S.C. §1001 (false statements to the federal government). This isn't just a platform matter, it's a legal record matter. Don't put that at risk.

colift doesn't pad hours, round up, or certify more than was measured. The same discipline applies to volunteers.