Can a company pay to rank higher on PickOneAI?

No. There is no paid placement. Payment, sponsorship and affiliate status are not inputs to the score.

How is the pick chosen?

Scores are recomputed daily from public signals. The current pick keeps its spot unless another tool's score beats it by at least 1.0 points; then the highest scorer becomes the pick. If a job has no previous pick, or the previous pick is no longer listed, the highest score wins outright. The next tool in line is the runner-up.

Why does a pick sometimes stay put when another tool scores slightly higher?

Because of our 1.0-point stability margin. Scores move a little every day, and without the margin a pick could flip back and forth on gaps as small as 0.1 points. We always show both scores, and we say so on the page when a runner-up leads on raw score but hasn't cleared the margin.

Home / Methodology

Methodology

Every pick comes from one formula, recomputed daily. We publish it so you can check our work and disagree with the inputs.

The formula

score = 100 × Σ(wᵢ × sᵢ) / Σ(wᵢ)      summed over the signals available for that tool

Fit        w = 0.4   s = (fit − 1) / 4                              editorial, 1–5
Adoption   w = 0.25   s = log10(1 + stars) / log10(1 + max stars)        GitHub stars, catalog-wide max
Momentum   w = 0.2   s = clamp(1 − days since last stable release / 90, 0, 1)
Attention  w = 0.15   s = log10(1 + HN points) / log10(1 + max HN points) last 30 days

Pick = the current pick, unless another tool's score beats it by ≥ 1.0 points;
       then the highest score. No previous pick (or it dropped out) → highest score.
Runner-up = next in line. Score ties → higher fit, then name.

Fit is our judgment of how well a tool does the specific job. It's the one human input, and we label it as editorial everywhere it appears.
Adoption uses GitHub stars from the tool's main public repo. Closed-source tools have no star signal. It's dropped, not set to zero, and the other weights are re-normalized.
Momentum uses the latest stable GitHub release (pre-releases are filtered out), or the last push if there are no releases.
Attention sums Hacker News points for stories from the last 30 days that link to the tool's site, or whose title names the tool when it has no unique domain. Title matching can over-count.
Star velocity (7-day change) is collected from daily snapshots and shown on tool pages. It isn't weighted yet. We'll add it once we have 30 days of history, and we'll announce the change here.

The stability margin

Scores are recomputed every day, and small daily moves in stars, releases and Hacker News points can shuffle tools that are nearly tied. Without a rule, a pick could flip back and forth on gaps as small as 0.1 points, which tells you nothing. So the current pick keeps its spot unless a challenger's score beats it by at least 1.0 points. Once a challenger clears that margin, the highest scorer becomes the pick. If a job has no previous pick, or the previous pick is no longer one of its candidates, the plain highest score wins.

The margin only decides when a pick changes. It never changes a score: every score shown is the raw formula output. The "previous pick" is the pick we last published, read from our own data before each daily recompute. When a runner-up leads on raw score but hasn't cleared the margin, we say so on that job's page and list it below rather than hide it.

Close calls right now (gap smaller than 1.0 points):

Job	Pick	Closest challenger	Gap	Status
Vector database	pgvector 77.8	Qdrant 77.7	0.1	pgvector leads on raw score

Data sources

GitHub repository stats via the GitHub REST API (or the public ungh.cc mirror of it), and GitHub releases Atom feeds. For tools with no public GitHub releases (for example GitHub Copilot, Vercel, Deepgram, Exa), the release feed also lists entries we check by hand against the vendor's official changelog, each linked to its source. Those entries are news only and never change a score.
Hacker News via the Algolia HN Search API.
Pricing read by hand from each vendor's pricing page. Every summary links its source and the date it was checked. If we can't read a price, we say so instead of guessing.
Product Hunt: we link to it but don't pull its data. Its API terms forbid commercial use without permission.

Signals last fetched Oct 21, 2026. Raw data: /api/picks.json and /api/tools.json.

The no-paid-placement rule

Nobody can pay to be a pick, a runner-up, or to move up a list. We don't sell rankings, "verified" badges, dofollow links or guest posts. If we ever sell anything to vendors (for example, a labeled "featured launch" in the release feed), it will be visually separate from rankings and will never feed into the score.

Affiliate disclosure

Some outbound links may become affiliate links, meaning we may earn a commission if you sign up. Affiliate links are marked affiliate and use rel="sponsored". Affiliate status is not an input to the formula and never changes a score or rank. We include tools with no affiliate program, and we pick them when they score higher. As of today, no links on this site are affiliate links.

Corrections

If a price, date or claim is wrong, the source link next to it shows where it came from. Every claim on the site has one. Report errors and we'll fix them and note the change.