Twitter should not only open source its algorithm but also create a marketplace where anyone can build their own algorithms, and use the ones created by others
increases usage — decreases assholery — rebuilds trust I wrote all about it how it could work and why to do it here: https://every.to/... @ejang : Open platform co-op https://twitter.com/... @hkanji : “Commoditize your complement” https://every.to/... via @getmatterapp Andrew Hutchinson / @adhutchinson : An interesting look at the proposed open source algorithm approach for Twitter from @nbashaw - it eyed certainly be worth exploring if Twitter's entity classification, in particular, could be improved to make its algorithm more effective https://every.to/... @theaksheffect : @nbashaw @Twitter This is actually what @jack wanted to do and mentioned it multiple times while he was CEO Carl Quintanilla / @carlquintanilla : “When Musk was pretending to take $TSLA private, he pretended that he was going to let all of Tesla's existing shareholders remain shareholders of the new ‘private’ Tesla. This is not a thing.. And now he's doing it again.. Sure buddy!” @matt_levine $TWTR https://www.bloomberg.com/...
Context & Ripple Effects
The proposal treats recommendation logic as a platform layer: Twitter would expose it and let outside developers build, share, and use alternative ranking systems. It frames user trust and lower-abuse incentives as reasons to move beyond a single company-controlled feed.
Related coverage later tests the premise. Twitter ultimately released part of its For You recommendation code, but researchers found that release lacked the context needed to explain recommendations, while experts had already argued that code alone cannot make a data-heavy recommendation system transparent.
First-order effects
- Twitter would shift from operating one recommendation system to governing a marketplace of third-party algorithms, with developers and users participating directly in feed selection.
- Users would gain a mechanism to choose among ranking approaches rather than accepting Twitter's default feed, making the platform's trust proposition depend on both disclosure and usable choice.
Second-order effects
- Third-party algorithm builders would compete on feed quality and safety, while Twitter would need rules for access, distribution, and accountability across externally built recommenders.
- A code release without the operational context identified in the later transparency critique would limit outside developers' ability to reproduce or meaningfully evaluate Twitter's recommendations.
Third-order effects
- If platforms pair openness with genuine developer access and user-level choice, recommendation systems can evolve from proprietary product features into governed marketplaces.
- The later partial release suggests the durable dividing line is not whether code is published, but whether the surrounding data, context, and governance make that openness usable.
The trend: Social platforms are being pushed to treat recommendation transparency as a combination of code access, operational context, and user choice rather than a source-code release alone.