- Roko's Basilisk
- Posts
- Siri Failed, The Hinge Didn't
Siri Failed, The Hinge Didn't
Plus: Trump dismisses AI critics, OpenAI delays going public, Altman slows the frontier.
Here's what's on our plate today:
🧪 Apple uses AI to match each iPhone Duo hinge with its frame.
📰 Trump dismisses AI critics; OpenAI's IPO slips past 2026; Altman backs pacing.
💡 Prompt of the Day: find the one narrow, measurable job AI can own.
Let’s dive in. No floaties needed.

Goodies delivered straight into your inbox.
Get the chance to peek inside founders and leaders’ brains and see how they think about going from zero to 1 and beyond.
Join thousands of weekly readers at Google, OpenAI, Stripe, TikTok, Sequoia, and more.
Check all the tools and more here, and outperform the competition.
*This is sponsored content

Business and tech news. One visual per story. Three minutes.
Bay Area Times is the free daily newsletter that gets business and tech leaders up to speed before their first meeting.
AI, startups, robotics, biotech, energy, and the wider innovation economy, each story paired with one clear visual so you see the point instead of digging for it.
Monday to Friday, three minutes, no noise. More than 250,000 founders, operators, and investors already read it.
*This is sponsored content

The Laboratory
TL;DR
Apple has struggled to get Siri right, yet it has found a job for AI inside the iPhone Duo's hinge.
The fitter: Ford removed the factory fitter by making parts precise, and Apple has brought one back as AI for its first foldable.
The match: Apple said on September 9, 2026, that AI pairs each hinge with its best-fit frame, while a laser and a printer handle the rest.
The contrast: the smarter Siri arrived only after years of delays, and Apple's next AI models will be built on Google's Gemini.
The narrow job: matching hinges has one measurable goal, which may suit AI better than open-ended tasks like answering whatever a person asks.
The stakes: reviews around the October 23 release, set against Samsung's Galaxy Z Fold 8, will show whether AI matching works or adds little.
Apple's foldable brings back the factory fitter, with AI picking the pairs
Early car factories built each car in one spot with a crew that included skilled workers called 'fitters'. They were needed because parts often came from machine shops slightly different from their drawings, so a fitter filed every piece until it sat properly against its neighbors. However, since every car needed that handwork, a factory could build only as many cars as its fitters could finish. When Henry Ford came up with the concept of the assembly line, he had to find a solution that would reduce and eventually eliminate the job of fitters, and the answer was to demand parts made so precisely that any one fit any Model T without filing, which took the fitter off the floor. The fitter had existed only because parts could not be made precise enough to go together on their own.
More than a century later, Apple is using AI as a fitter on its first foldable phone, the iPhone Duo. When the company unveiled the Duo on September 9, 2026, Johny Srouji, its chief hardware officer, told the audience, "During manufacturing, we use AI algorithms to precisely match each individual hinge with its best-fit housing.” The housing is the frame the hinge mounts into on either side of the fold, and choosing which hinge suits which frame is the kind of fitting Ford's precise parts had made unnecessary.
Phones are normally built the way Ford built cars: every part is made to fit any unit on the line. However, since a folding phone's screen must lie flat over the hinge it bends around, tiny differences between the hinge's many parts add up and can leave the display uneven. Apple's answer is to check every Duo individually, with AI choosing the pairs, a laser scanning each finished surface, and a printer adding thin layers to smooth what remains.
The AI fitter matches parts that are already made
Matching starts with measurement, because every hinge and every housing leaves its production line a fraction different from its design. Apple's AI then picks, for each hinge, the housing whose small errors best offset its own, so the pair lands closer to the intended shape than either part would with a random partner. Engineers call this 'selective assembly', a technique studied since at least the 1980s that lets a factory build an accurate product from parts that are individually less precise. For the Duo, a better match means a hinge that sits evenly in its frame, which gives the screen above it a flatter base.
Srouji's second step, which he did not attribute to AI, scans each finished unit with a focused laser that maps its surface height point by point. A printer then adds up to 25 micro layers of a custom liquid plastic that hardens under light, removing the small ripples that pairing alone could not. The AI chooses the pairs, while the scanner and printer do the measuring and smoothing that make up the rest of the fitter's work.
Apple has not said what kind of AI makes that choice, whether software that learns from production data or the established math researchers have applied to such pairings for decades. TechCrunch's headline said the hinge was built with AI, and its opening line said AI had helped design it too. Srouji described only the factory, and Apple's press release and product page never mention AI in how the hinge is made. The AI works on parts that already exist, which makes its job fitting in the plainest sense.
The factory gives Apple an AI problem narrow enough to solve
Siri, the AI product most people associate with Apple, has been a much harder assignment. The smarter Siri Apple promised arrived only after years of delays, and in January Apple and Google said the next generation of Apple's AI models would be built on Google's Gemini. The company that needed outside help to rebuild its assistant is finding its own practical use for AI in the making of a phone.
Apple has not said why AI went into the hinge line, but the fitter's job has something Siri lacks: a definite end point. The AI has one task, choosing the housing each hinge goes into, and one test of success, a pair that lands as close to the design as the two parts allow. Each choice can be measured against the finished phone, so there is no argument about whether the software got it right. A bad pairing is caught at inspection long before a customer opens the box.
The hinge may therefore be an appropriate job for AI, since such a definite goal makes success far more likely. Open-ended goals are where AI tends to struggle most, because there is no single right answer to aim for or to check against. Siri has to be ready for almost anything a person asks, while the fitter answers one measurable question over and over.
Whether the approach produces a flatter, sturdier phone is still unproven, and early impressions are mixed. MacRumors, after handling the Duo, found the crease, the line where the screen folds, visible in certain lighting but better hidden than on Samsung's foldables. 9to5Google, reviewing hands-on footage, judged it as noticeable as the one on the Galaxy Z Fold 8. TechCrunch's Lucas Ropek wrote that it is too early to know whether the hinge will hold up, and Apple publishes no rating for how many folds it should survive.
Ford could take the fitter off the floor because his factory learned to make parts precise enough that they needed no filing. Apple has not reached that point with the Duo's hinge, so it has rebuilt the fitter as software and machinery that work on every phone. Reviews around the Duo's October 23 release will show whether that narrow use of AI works or hardly makes a difference. Samsung credits a new material it calls Flex Titanium with reducing the crease on its Galaxy Z Fold 8. If the Duo looks and folds no better than the Fold 8, the AI fitter may be adding very little.


Prompt of the Day
![]() | 💡Act as an operations analyst. Look at a workflow I describe, find the narrowest task inside it that has one measurable right answer, and specify exactly what data would prove the AI got it right every single time. |

Hire smarter with Athyna, save up to 70% on salary costs.
Athyna connects you with top LATAM AI talent, fast!
Meet vetted professionals in as little as five days, without long, expensive recruiting cycles.
Save up to 70% on salary costs when hiring AI engineers, product leaders, and data scientists.
Get AI-assisted matching plus human vetting, so your shortlist is tight, and your interviews are worth it.
*This is sponsored content

Bite-Sized Brains
Trump waves off the doomers: The president called AI's critics "very negative forces" raising things that won't happen, adding that whoever wins AI wins.
OpenAI's IPO slips past 2026: Altman told Fortune that listing now would be ill-advised given safety concerns, and that even a 10% extinction risk would be unacceptable.
Altman backs pacing the frontier: The OpenAI chief endorsed Amodei's language on pacing, saying the shift to safety carries a cost but keeps confidence in US labs intact.

Tuesday Poll
🧪 Apple put AI on the hinge line, not inside Siri. What does that tell us? |
The Toolkit
Chroma: Open-source vector database for AI apps, quick to set up and easy to scale for RAG.
Lavender: AI sales email coach that scores your drafts live and suggests rewrites to lift reply rates.
VEED: Browser-based video editor with AI subtitles, voice cloning, and one-click translation for fast social edits.

Rate This Edition
What did you think of today's email? |






