Oct 28, 2026 · 2h 7m
Apple’s hardware edge meets the bottlenecks of AI’s next phase
Apple Pro and Tell, Genie 4, The Gurmanator Joins | Mark Gurman, Zach Rash, Ben Liu, Amiram Shachar, Jasper Zhang
The episode connects Apple’s product strategy with the infrastructure, security, robotics, and drug-development systems needed to turn AI ambition into durable businesses.
- 1Apple’s foldable iPhone and touchscreen MacBook Pro expose a familiar tradeoff between visible hardware novelty and slower software progress.
- 2AI infrastructure is becoming a market of its own, spanning gigawatt-scale demand, cloud security, fragmented GPUs, and uncertain utilization.
- 3Coco Robotics, Formation Bio, Upwind, and Hyperbolic show how startups are applying AI to physical logistics, medicine, security, and compute access.
Don't miss
The hosts throw the iPhone Duo across the studio while Mark Gurman discusses its cracked-display controversy and durability concerns.
The brief
The Apple discussion starts with the touchscreen MacBook Pro and foldable iPhone Duo, but lands on a sharper distinction: OLED improves the laptop more convincingly than touch does.
Mark Gurman examines the iPhone Duo’s durability, software quirks, Apple Intelligence, smart-home ambitions, and rumored glasses, while the hosts weigh Apple’s hardware strength against its software lag.
The conversation widens from products to infrastructure: AI demand, model naming, neocloud economics, security risks from agent-written code, and the challenge of making compute capacity useful and accessible.
Four founders map AI’s expansion into the physical and operational world: Coco scales sidewalk delivery, Formation Bio targets clinical bottlenecks, Upwind secures cloud systems, and Hyperbolic aggregates GPUs.
The episode’s central tension is execution: Apple must turn ecosystem control into better software, while startups must make ambitious AI systems reliable, regulated, and economically viable.
What was said on this episode
75 statements · 45 positive · 17 negative · 4 mixed · 9 neutral
OLED makes sense as a MacBook Pro upgrade.
“bringing that to the MacBook Pro makes a lot of sense.”
Listen at 5:07
OLED benefits video editors and color-correction professionals.
“I do think video editors, anyone doing important graphics or color correction, especially if they're delivering to an OLED TV.”
Listen at 5:30
Apple logo illumination likely still costs about the same per unit.
“the technology that it costs to light up the Apple logo probably hasn't changed in price.”
Listen at 15:36
Apple should preserve premium branding through aesthetic details.
“I think that's the premium Apple brand that they need to stick with and stay with and stuff.”
Listen at 16:08
Genie 4 will enable interesting real-time world-model applications.
“So there will be some interesting like real time applications as this rolls out.”
Listen at 18:44
Model naming should become task-oriented rather than version-number based.
“I hope that the version numbers kind of melt away and you're just in the world of like, okay, like... Big model for this task, small model for that task.”
Listen at 25:17
Expanding AI demand will benefit many semiconductor and cloud providers.
“rising tide will like lift all boats as everything becomes”
Listen at 28:46
Different AI chips will find optimal roles across inference workloads.
“there will be an optimal way to. to get a model running on that.”
Listen at 29:09
Seven-year-old A100 hardware has depreciated but retains value.
“there is depreciation, but there's still some value there.”
Listen at 32:35
Malaysia functions as a strategic intermediary for AI compute.
“They are the Switzerland of AI.”
Listen at 34:19
TurboPuffer prioritizes minimizing latency in knowledge retrieval.
“every millisecond matters, and that's what the TurboPuffer team prioritizes.”
Listen at 41:47
TurboPuffer exceeds $100 million ARR with roughly 50 employees.
“now they're doing over $100 million in ARR with roughly 50 employees.”
Listen at 41:52
Private investors will capture a disproportionate share of wealth creation.
“Wealth creation is going to be heavily skewed towards private investors over the coming decade.”
Listen at 42:43
Frontier AI capabilities require marketing to achieve adoption.
“unless there is a commensurate marketing push, it's not guaranteed that it gets adoption.”
Listen at 45:45
Google is slower than small AI-agent startups at product execution.
“they're always a little bit slower on the product side than like an instinct, which is just a couple people moving really fast.”
Listen at 47:34
Apple or Amazon could acquire Instinct at its high valuation.
“it could happen if Apple or Amazon make a crazy move.”
Listen at 49:33
An Apple or Amazon acquisition of Instinct should receive antitrust approval.
“It should 100% be allowed.”
Listen at 49:46
Personal AI agents are currently an exceptionally competitive technology category.
“this is the most competitive tech category we've had in years.”
Listen at 50:27
Instinct may become a major personal-AI-agent company.
“maybe the market's so big that they just become a really big player in personal ai agents”
Listen at 52:38
iPhone Duo display cracks are caused by a hinge-component manufacturing defect.
“the cause of this is a manufacturing defect related to one of the components in the hinge”
Listen at 57:10
iPhone Duo owners cannot obtain replacement devices from Apple stores.
“You can't get a replacement.”
Listen at 59:09
The $80 Apple standard case provides a good iPhone Duo experience.
“I've had great experience with the cheaper $80 Apple standard case.”
Listen at 1:01:59
Initial iPhone Duo app-optimization complaints are no longer significant.
“The app thing is a non-issue.”
Listen at 1:03:14
iPhone Duo software quirks remain its main problem.
“The real problem with the iPhone Duo is there are some software quirks”
Listen at 1:03:57
Apple should accelerate its software-update cadence.
“they need to shake that up.”
Listen at 1:04:15
The foldable iPhone Duo is a highly successful product.
“I would never use a non-foldable iPhone at this point. I'm completely sold on the bigger screen. I think it's absolutely home run product.”
Listen at 1:05:50
Gurman uses the iPhone Duo inner screen about 75% of the time.
“for most of the time, 75% of the time, I'm using the inner screen.”
Listen at 1:06:18
The iPhone Duo display-cracking hardware issue is genuine.
“this hardware issue with the crack display is a real thing.”
Listen at 1:06:36
The new MacBook Pro weighs four pounds after internal redesign.
“The new one looks identical, but they reworked the internal, so now it's four pounds.”
Listen at 1:07:49
Apple will release M7 Pro and M7 Max MacBook chips late next year.
“There will be new M7 Pro and M7 Max versions at the tail end of next year.”
Listen at 1:09:29
MacBook Pro touchscreen functionality will see limited use.
“I think that it's going to be little used in the end.”
Listen at 1:11:45
Touchscreens are now expected features on premium laptops.
“It is table stakes. It is spec sheet 101. It is expected.”
Listen at 1:11:58
OLED is a stronger MacBook Pro selling point than touch.
“I think the OLED will look better. So that's the selling point.”
Listen at 1:13:11
Apple will prioritize Home View over a privacy-focused smart-home ecosystem.
“they're still not going to do it. They're going to do a home view, which no one ever asked for or needed. It has no real purpose.”
Listen at 1:13:54
Apple-branded smart-home accessories would be expensive and struggle for penetration.
“if Apple starts coming out with this stuff and self-branding it, they're going to be at a very high price point. and they're not going to be able to get market penetration.”
Listen at 1:14:49
Apple smart glasses will be a major upcoming product.
“The smart glasses are the big things for next year.”
Listen at 1:19:17
Apple smart glasses will function like AirPods in glasses form.
“It's AirPods in a glasses form. It's AirPods for your face.”
Listen at 1:19:32
Apple smart glasses will compete with Meta glasses.
“It's a competitive product to the metaglasses.”
Listen at 1:19:45
Meta’s Luxottica partnership improves glasses appeal because consumers reject Meta branding.
“That was super smart on Meta's part to partner with Luxottica because nobody would buy Meta-branded glasses.”
Listen at 1:20:04
The MacBook’s tandem OLED panel may enable the illuminated logo.
“using the tandem OLED panel, so the thinner panel. Yeah. And the brightness benefits. you're able to use the display, the back of the panel as a way to light it up.”
Listen at 1:22:25
The 2021 MacBook Pro design fixed the product’s major shortcomings.
“With this one that you've got, the 2021 MacBook Pro design, they fixed the MacBook Pro. This is the perfect MacBook Pro.”
Listen at 1:23:18
Apple now has a better chance of acquiring Instinct than previously.
“I think they have a better chance of buying Instinct than they did.”
Listen at 1:26:24
Apple’s software is currently its weakest area.
“The software. side of things right now i think is the weakest part of apple”
Listen at 1:26:34
Apple earns more from casino-style games than from the Mac business.
“Apple makes more money from casino games, that type of game, than they make on the Mac.”
Listen at 1:27:11
- Zach Rashon Coco delivery robotsNeutral1:32:13
Coco delivery robots operate as a shared fleet.
“they're shared.”
Listen at 1:32:13
- Zach Rashon Coco delivery robotsPositive1:33:15
Coco robots can operate about 20 hours daily with four hours downtime.
“So you run them 20 hours a day, and then maybe you have four hours of downtime to do cleaning and charging.”
Listen at 1:33:15
- Zach Rashon Coco delivery robotsPositive1:34:07
Coco expects to operate in 40–50 cities next year.
“next year we should be in between 40 to 50 cities across even more continents.”
Listen at 1:34:07
- Zach Rashon Coco delivery robotsPositive1:34:56
Coco’s fleet will eventually reach one million robots.
“the path to a million feels pretty inevitable from there.”
Listen at 1:34:56
- Zach Rashon City sidewalksNeutral1:37:09
City sidewalks are the world’s most complex dynamic non-rule-based environment.
“the most complex dynamic not rule-based environment in the world.”
Listen at 1:37:09
- Zach Rashon Sidewalk delivery robotsNeutral1:37:29
Sidewalk robots must learn socially appropriate navigation in shared public spaces.
“the robot actually needs to learn how to be a great user of the shared space with the public.”
Listen at 1:37:29
- Zach Rashon Coco robotics datasetPositive1:38:01
Coco likely has one of the largest sidewalk-robotics datasets.
“we probably have one of the largest data sets of this sort of data.”
Listen at 1:38:01
Delivery robots should prioritize being helpful without annoying pedestrians.
“making sure that friendly robots are never annoying has to be number one priority.”
Listen at 1:38:52
- Zach Rashon Research monkeysPositive1:41:06
Zach Rash recommends increasing the number of research monkeys for national security.
“There is. They're important. National security. We've got to increase the number of monkeys we have.”
Listen at 1:41:06
- Zach Rashon Discovered drug candidatesPositive1:41:35
The number of discovered drug candidates doubled over the previous decade before AI.
“there's been a doubling of the number of discovered drug candidates in the past 10 years, even before AI.”
Listen at 1:41:35
- Zach Rashon Approved drugsNeutral1:41:39
Annual approved-drug counts have remained roughly constant at 50.
“The number of approved drugs has kept, like, kind of constant at 50 a year.”
Listen at 1:41:39
Clinical trials can cost hundreds of millions; Phase 2 trials can exceed $60–80 million.
“a clinical trial can cost a few hundred million dollars. Like a phase two clinical trial can be $60 million to $80 million, if not more.”
Listen at 1:42:59
- Zach Rashon Drug design and human clinical validationMixed1:43:40
Drug design is becoming commoditized, while human efficacy validation remains scarce and unchanged.
“design is getting increasingly commoditized. But what is pretty rare is like the ability to prove that it works in a human. And that capacity hasn't changed.”
Listen at 1:43:40
A full Alzheimer’s clinical program can cost approximately $500 million.
“When a single clinical trial in Alzheimer's across all the phases can cost like hundreds of millions of dollars, actually it's half a billion per shot.”
Listen at 1:45:57
- Zach Rashon AI drug discoveryPositive1:46:09
AI will likely increase drug-discovery output by up to roughly 100-fold.
“And now with AI, that's probably going to tend. if not 100x, right?”
Listen at 1:46:09
- Zach Rashon Post-Phase-2 drug assetsPositive1:47:20
Post-Phase-2 clinical proof-of-concept drugs are valued at roughly one times peak sales.
“Once it has clinical proof of concept, so in human trials, in the disease, post phase two, the heuristic is one X peak sales.”
Listen at 1:47:20
- Zach Rashon Post-Phase-3 drug assetsPositive1:47:31
Drugs past Phase 3 are valued at roughly two to three times peak sales.
“Once it's over phase three, past phase three, it's two to three X peak sales.”
Listen at 1:47:31
- Zach Rashon Phase 2 clinical trialsNegative1:47:57
Phase 2 trials succeed only about 30% of the time on average.
“Phase two is only worth 30% of the time on average.”
Listen at 1:47:57
- Zach Rashon Phase 3 clinical trialsMixed1:48:13
After a successful Phase 2, Phase 3 has a 40% failure and 60% success probability.
“There's a 40% chance it goes to zero. A 60% chance to phase three works.”
Listen at 1:48:13
- Zach Rashon Drug-development portfolioPositive1:49:26
A portfolio of ten drug-development shots is expected to produce three successes.
“Once you run 10 shots, you expect 3 out of 10.”
Listen at 1:49:26
- Zach Rashon Formation Bio fundPositive1:49:42
The fund’s diversified financing model lowers Formation Bio’s cost of capital over time.
“it lowers your cost of capital over time.”
Listen at 1:49:42
- Zach Rashon Public opinion of pharmaceutical companiesPositive1:50:26
Zach Rash predicts public opinion of pharma will improve as longevity drugs advance.
“I'm very optimistic it's going to be a lot better specifically because I think we're beginning to pursue drugs that just help you live longer and healthier.”
Listen at 1:50:26
- Zach Rashon Longevity drugsPositive1:50:42
Longevity and healthspan-enhancing drugs remain at an early stage.
“we're in early innings of things that just help you live longer and healthier.”
Listen at 1:50:42
- Zach Rashon AI-enabled pharmaceutical developmentPositive1:51:32
AI systems could let pharma companies develop many drugs with hundreds rather than 100,000 employees.
“you can prosecute 10, 15, 20, 50 drugs and not have to be a 100,000 person company versus a few hundred people.”
Listen at 1:51:32
- Zach Rashon Research monkeys versus BitcoinPositive1:52:01
Zach Rash predicts research monkeys will ultimately outperform Bitcoin financially.
“I think the monkeys will ultimately translate into more.”
Listen at 1:52:01
Open-source AI models make harmful cyber activity accessible to nearly everyone.
“every person is a hacker with open source models, the damage you can create”
Listen at 1:54:02
- Zach Rashon AI and cybersecurity threatsNegative1:54:24
AI is intensifying cybersecurity threats.
“AI just makes it stimulating it.”
Listen at 1:54:24
- Zach Rashon AI-assisted codingMixed1:57:50
AI-assisted coding reduced coding time 90% while increasing review time 600%.
“my coding... Time has dropped down by 90%, but the review time is up by 600%.”
Listen at 1:57:50
- Zach Rashon AI-generated codeNegative1:58:01
AI-generated code requires especially careful review.
“you have to be really, really smart on how to review that.”
Listen at 1:58:01
- Zach Rashon Neocloud GPU revenuePositive2:02:20
Neocloud operators should target $10–$20 in revenue per GPU hour.
“If you're on NeoCloud, you want to maximize your revenue to $10, $20 per GPU hour.”
Listen at 2:02:20
- Zach Rashon Neocloud marketNegative2:03:11
The neocloud market has more than 300 providers and is highly fragmented.
“there are more than 300 plus NeoClouds. The market is super fragmented.”
Listen at 2:03:11
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.
Weekly email
State of the Pod, every Monday
The week's most-discussed names and stories from podcasts, with the moments behind them. Free, five minutes, unsubscribe in one click.