How our rankings work

How it works, in brief

We track strictly AI-native tools — products that would not exist if you took the AI away. Every night, an automated pipeline measures the public activity around each tool and turns it into two things you can see: a score, which is how much measured activity surrounds the tool right now, and a direction badge, which shows whether that activity is accelerating or fading over the trailing seven days. Within a category, tools are ranked by score.

Which tools get measured is a human, editorial decision — junk, vaporware, and thin wrappers don’t get in. What the measurements say is automatic. Nothing in between can be bought.

What we measure

Today, scoring uses two public, developer-centric sources, named so you can judge both the evidence and its limits:

• Hacker News — the discussion board where developers and early adopters surface and argue about new tools. Through its public search API we count the stories, points, and comments that reference each tool.
• GitHub — where much of the actual building happens. For tools with public code repositories we track stars and how quickly they’re being added, along with related activity such as forks.

If we ever add a source to scoring, this page changes the same day.

Matching a mention to a tool. Each tracked tool has a maintained set of search terms, domains, and — where the tool is open-source — one flagship repository. Ambiguous or generic product names are disambiguated by hand, and matches that can’t be made precise are excluded rather than counted. We would rather miss a mention than count someone else’s.

When a note on this site cites numbers — “14 stories, 37 points, 4 comments” or “65,461 stars” — these are those counts, always from public data.

Reading the numbers

Plain-English definitions — including what each number doesn’t tell you.

StoryHacker News

A post or link submitted to the community that references the tool.

How to read it: Coverage — the tool is being noticed. A story count alone says nothing about whether the reception was warm.

PointHacker News

An upvote on one of those stories.

How to read it: Readers found the story worth attention. That's interest in the story — not approval of the product.

CommentHacker News

A reply in a story's discussion thread.

How to read it: Conversation volume. Heavy commenting can be excitement, criticism, or a heated argument — it measures how much people are talking, not what they concluded.

StarGitHub

A developer bookmarked the tool's code repository.

How to read it: Usually a sign of interest — but not proof of use or endorsement. People star what they want to remember, compare, or keep an eye on.

ForkGitHub

A developer made their own working copy of the code to build on or modify.

How to read it: Hands-on engagement — a stronger commitment than a star. Usually adoption; either way, it's real effort spent on the tool.

Score and direction — what the number and the badge each tell you

Score measures current activity. The badge shows direction. The score reflects how much measured activity surrounds a tool right now — and nothing else. Whether that activity is accelerating or fading is reported separately, as the badge; it is not folded into the number. Rank within a category follows score.

Direction is measured against the tool’s own past, not against other tools. We compare the trailing seven days with that tool’s own recent history, and we publish the result as a direction rather than a number, because that’s the reading we trust:

• ↗ Rising — attention is accelerating.
• → Steady — activity is holding its level.
• ↘ Cooling — attention is fading from its recent level.
• Watchlist— measured nightly, but not enough history yet for a direction. Where there isn’t enough signal for a meaningful score either, we withhold the number rather than print a floor. New admissions start here; they move up the moment the data does.

This is why the two readings can seem to disagree: a tool can sit near the top on score while its badge cools, or rank mid-pack while rising fast. That’s not a bug — one tells you roughly where a tool is, the other tells you where it’s headed.

Attention is not endorsement

Our rankings measure the direction and intensity of real attention. They are not reviews, and a high rank is not a recommendation.

Spikes often follow launches, major releases, or a surge of interest. But attention can also spike for unflattering reasons: an outage, a pricing backlash, a security incident. We handle that three ways:

• A human gate at the door. Every tool passes editorial review before it enters the catalog, so the list starts from products worth measuring.
• The note says what moved.The “Why it’s rising” / “Why it’s cooling” note beside each tool summarizes what the data shows. When a move is driven by controversy rather than adoption, the note should say so — if you catch one that doesn’t, tell us.
• Both directions, plainly. Cooling is reported as visibly as rising. What we will never do is quietly re-rank a tool because its numbers got uncomfortable. We report measured movement consistently, even when the attention is negative.

What these rankings are not

Not revenue. Not user counts. Not a product-quality or safety rating. Not advice. A tool can be excellent and quiet, or mediocre and loud — rankings tell you where developer attention is moving, and that’s all they claim. Use them to decide what’s worth your own evaluation, not to skip it.

Where our sources fall short

Hacker News and GitHub are developer-centric. Categories closest to developers — code assistants, say — carry the strongest signal. Consumer-leaning categories are seen through a developer lens: a tool that classrooms love but developers don’t discuss will under-index here. We’d rather say that plainly than pretend coverage is uniform, and it’s one reason we add sources carefully rather than quickly.

Independence

No paid placement. No pay-to-rank. Ever.

The site will be supported by clearly labeled ads and affiliate links, never by rankings. An affiliate relationship never touches a score, a rank, or a badge, and it will always be disclosed where it appears. If a vendor asks how to improve their rank, the honest answer is the only answer: ship something developers talk about.

What we don’t publish

We keep exact weights private to make targeted manipulation harder — and publish the sources, definitions, time windows, and limits you need to interpret the rankings. The full-resolution history behind these rankings — every tool, every night — is reserved for a forthcoming data product for teams; the public site always shows the honest direction.

See something wrong?

Corrections are part of the method. Write to info@bestofnow.com and we’ll check the data.

Change log

When the instrument changes, it gets recorded here — so anyone who cited a ranking can see whether the ground moved underneath it.

August 21, 2026Direction no longer affects score

Score now measures current activity alone. Direction is still measured and still published as the badge, but it no longer adjusts the number or the rank. Displayed scores shifted once as a result.

Why: Direction and level answer different questions, and blending them let short-lived swings move ranks further than the underlying activity justified. Separating them lets each reading say only what it can support. A fuller revision of how direction is measured is in testing, with a decision expected in September 2026.

About bestofnow →