AI NewsGo-to-marketReported
Synth makers say Apple's app review now hallucinates fake rules
Synth app makers say Apple's App Store reviewer is now rejecting apps for rules that do not exist. One asked for a license for a fake audio format called Polyharmonic.
Image: Synth Anatomy
Why it mattersIf a model is grading a submission without telling you, you have to write for it. Documentation and clear scope now matter more than they did last month.
A software team can release a product and follow every App Store rule. The reviewer can still ask for a license on something that does not exist. That is what Michael Hetrick says happened to Unfiltered Audio. His synth app, Sinusoidal Run Rhythm, was rejected by Apple. The reviewer wrote that Hetrick had to show a license for Polyharmonic. There is no audio technology called Polyharmonic. The reviewer, in his account, invented it.
Hetrick's account is quoted in Synth Anatomy by the writer Tom. The piece collects complaints from audio software makers about Apple's review. Synth Anatomy says review has turned into a long wait and a wall of rejections. Hetrick told the site the back-and-forth stretched over a month. He released the app with no audio in the end, just to get it on the store.
What other developers told the site
Pittsburgh Modular's Microvolt 3901 came out as a desktop plugin, per Synth Anatomy. It reached iOS more than two weeks later. AudioThing's Materia took 25 days to pass review and had one rejection in that time, per the site. Its Seconds app, made with the artist Hainbach, had been in review 29 days as the story was published. Igor Vasiliev's HollowTone sat in review for weeks.
Marek Dendes, who makes the CRUX Synth, said his app was rejected four times in a row. Each rejection was called "design spam." Synth Anatomy quotes him: the status goes from waiting for review to rejected instantly, he said, so nobody even reads what he writes.
What the piece infers, and what it does not
Synth Anatomy says the pattern points at an AI-grading step. Apple has not said anything about that publicly. The site is careful about this. It quotes developers. It does not say Apple has confirmed. The strongest piece of evidence it holds up is the Polyharmonic rejection. A human reviewer would not invent a protocol and then ask for a license for it. A model asked to check a description against a rule set might.
The discussion ran on Hacker News with 128 points as of writing.
What a buyer and a founder can take away
The reader here is anyone whose product reaches an iPhone through the App Store. A reviewer who is in part a model has two effects. A vague description, or a scope that sounds off-script to a model, is now likelier to get rejected. You also cannot appeal to a model the way you can to a reviewer at a desk. A clearer, more conservative description costs you nothing. Documented scope that matches the binary may save you weeks.
For an AI-native team, there is a second lesson. If you put a model into a review loop in your own product, this is what breaks. A model will refuse a request for the wrong reason unless something else is checking its work. An audit trail of what the model decided and why helps. A second human review on a refusal helps too. Both are cheap ways to keep the Polyharmonic case out of your own product.
Source
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.
