The formula, printed.
Most platforms describe their ranking in adjectives. These are the numbers Surf multiplies by, the controls that move them, and the switch that turns the whole thing off.
Algorithm transparency usually arrives as a blog post explaining that a feed considers relevance, recency and relationships. That is a description, not a disclosure: it cannot be checked, argued with, or acted on. The useful version is narrower. Publish the weights. Put the controls that move them in the product. Let any post say which signals surfaced it.
What each signal is worth.
23 named signals feed the score. These are the ones that move it most. Engagement is weighted by how much work the action takes, and anything you say explicitly outweighs all of it.
| Signal | Weight | Why |
|---|---|---|
| Reply | 7.0 | The heaviest engagement signal. Conversation costs the most effort |
| Repost | 5.0 | Putting your name on someone else's post |
| Quote | 5.0 | A repost that adds something |
| Like | 0.5 | Cheap to give, so worth little |
| Show more | +15 | You asked for it explicitly |
| Show less | -20 | You asked against it explicitly |
| Not interested | -50 | The largest single weight in the system |
Ten sliders that move the real weights.
Three set the balance of the feed, seven set the character of what fills it. They scale the numbers the ranker multiplies by, rather than filtering the result after it has been ranked. Behind them sit 8 topics and 49 sub-topics.
See every controlOr write it as a sentence.
Up to twelve standing rules in plain English, 120 characters each. Less negativity, more from people I know. Each one stays visible, each one has a switch, and anything you set by hand beats what the system inferred about you.
See all featuresDescribe your feed
Every post says why it is there.
One tap reads the reason straight off the ranking query that surfaced the post, so it is the actual signal rather than an explanation written afterwards. No second request, and no model in the loop.
How ranking worksThe water was unreal this morning.
Why am I seeing this?
Gaining traction right now, from a topic you lean into.
What this does not buy you.
- Published weights are not the same as a reproducible feed. Your candidate set depends on who you follow and what you have done, so two accounts with identical settings still see different feeds.
- The weights can change as the product changes. What is promised is that the page keeps up with them, not that they are frozen.
- Knowing the formula does not make ranking neutral. Any ordering is a set of choices; publishing them makes the choices arguable rather than invisible.
- A subscription multiplies the reach of a paying account's own posts, 1.5x to 4x. It is inside the published formula rather than outside it, but it is still an advantage money buys.
Questions people ask.
- What is algorithm transparency?
- In its weak form it means a platform describes, in general terms, what its ranking pays attention to. In the form worth having it means the actual weights are published, the controls that move them are in the product, and any post can tell you which signals surfaced it. Surf publishes 23 named signals with their weights, exposes them as ten sliders, and answers 'why am I seeing this' on every ranked post.
- Is there a social network where you control the algorithm?
- Surf is built around it. Three balance sliders and seven vibe sliders map onto the real scoring weights rather than filtering the result afterwards. Behind those sit 8 topics and 49 sub-topics, and up to 12 standing rules written in plain English at 120 characters each. A setting you make by hand overrides what the system inferred about you.
- Can you turn a social media algorithm off?
- On Surf, yes, with one switch. The feed drops to newest first, with no scoring and no learning from what you do. Blocks, mutes and comfort filters still apply, because those are settings you chose rather than ranking decisions made for you.
- How do you know why a post is in your feed?
- Any ranked post answers it in one tap. The reason is read straight off the ranking query that surfaced the post, so it reflects the actual signals rather than a plausible explanation generated after the fact. No extra request, and no model involved. Show less sits in the same sheet.
- Does publishing the formula let people game it?
- Somewhat, and it is a fair objection. The counterweights are that explicit feedback outweighs engagement by a wide margin, that no more than three posts from one author run before the rest are damped, and that any single format taking over a slate is damped too. The judgement is that a formula people can argue with is worth more than one nobody can inspect.
More on the algorithm, ad-free social media, and what Surf is.