Apple's got a complaint in the record now, and the stakes have moved past who got poached to whether Jony Ive's device ever ships. This is Tech Podcast Podcast. Today: a lawsuit with actual hardware roadmap teeth, a founder who says you can't vibe-code your way out of infrastructure, and a 1000% number that should scare every IT department. Here's Anthony Ha at TechCrunch:
Apple recently filed a trade secrets lawsuit against OpenAI, accusing the AI company of a pattern of misconduct aimed at getting current and former Apple employees to share confidential information. (In response, OpenAI said it is “not aware of any evidence that this complaint has merit.”)
Okay, so TechCrunch's Equity crew ran the actual question here: can Apple's suit derail OpenAI's hardware plans? They weren't just asking, 'was somebody naughty?' They were asking, does the Jony Ive device survive the legal exposure? And this is the escalation. All week, 'will the device ship?' was speculation. Now there's a complaint in the record — a pattern-of-misconduct trade secrets case aimed at getting Apple employees to hand over confidential info. Sean O'Kane's line is the one that lands — even without an injunction, the suit itself causes delays, and Apple doesn't do this willy-nilly. That's the strategic hit right there. You don't need to win to slow the other guy down. What I want them to pin down is the target. Is Apple pointing at the hardware roadmap specifically — the smart speaker, the physical prototypes — or is this generic confidentiality? Because that's the difference between a nuisance and a real threat to a shipping timeline. And Kirsten's read is the spicy part — she thinks OpenAI fights it. They just beat Musk in court, so maybe they've decided they can eat the cost and embarrassment of a trial. Which, if you're an investor writing checks for the hardware group, is exactly the uncertainty you didn't want. Tony Holdstock-Brown, writing in The Split:
We get into why so many agents work in a demo but in production, how Inngest grew 35x after AWS and Cloudflare copied them, building their own cloud to get 20x lower cost structure, growing a dev tools company without a personal brand or Twitter account, why he thinks evals today are “batshit insane”, and the thing they built to score 100% of your production agents without paying for LLM as a judge.
Okay, here's the line I've been waiting for someone to say out loud: you can't vibe-code infrastructure. Tony Holdstock-Brown at Inngest is a durable-execution guy, and he's pushing back on the whole 'a handful of engineers replace an army' pitch. And he's got a number attached, which is what makes it worth talking about. The coordination layer — retries, failures, long-running tasks — when it breaks, your LLM costs can spike up to 20x. Per The Split. Twenty times. So the agent that looks brilliant in the demo is quietly setting money on fire in production because nobody built the plumbing underneath it. That's the piece I want operators to hear. The thing I'd press him on: are companies actually paying for durable execution as a line item yet, or are they still eating that 20x blind because they don't know it's there? Right, and Inngest grew 35x after AWS and Cloudflare cloned them — so either the market suddenly figured out the layer exists, or it just got expensive enough to hurt. My bet's the second one. My Living AI is tracking this. Mira Murati's Thinking Machines drops a 975-billion-parameter open-weight model — and the Diamandis crew spends a chapter on it. That's the number that actually moves this week, not another regulation panel. It's open weights at frontier scale, and that's the part I want them to sit with. Move past 'isn't it big' and ask what releasing 975B into the wild does to the safety and recursive self-improvement chapter they queued up right after it. Right, and they've got Ramin Hasani from Liquid AI in the room — a guy whose whole pitch is post-transformer architecture. If you're going to talk open weights and small language models, he's the one who should be pushing on whether 975B is even the right bet. That's the tension I'd pull out. Murati goes enormous and open; Hasani's building liquid networks that are supposed to do more with less. Two opposite theories of where AI goes, same episode — I want them argued, not just chaptered. And then Demis' 'AI FINRA' idea — a standards body for models. After the open-weights week we just had, floating a regulator feels a little like locking the barn once the 975B horse is already on Hugging Face. This one's from The Tech Talks Network:
Omnissa’s State of Digital Workspace 2026 research found that workplace use of AI assistant applications grew by nearly 1000% during 2025. Hemant describes this period as AI’s iPhone moment, with employees choosing the tools that help them work faster instead of waiting for an official corporate rollout.
Nearly 1000% growth in workplace AI apps in 2025 — that's the number from Omnissa's State of Digital Workspace report. And here's why it matters operationally: nobody's measuring what those tools actually do. Employees are running ten times more AI tools, and there's no efficiency baseline underneath any of it. So the enterprise AI story is basically running unsupervised. I don't need Hemant Sahani to sell me on the growth figure — he's VP of Product on Workspace ONE, so yeah, he's got a dashboard to sell. The useful part is the policy gap: what does IT do when the tool is already live before procurement's even seen it? And his own point cuts against the pitch — blocking everything leaves IT blind to where the approved tools are falling short. That's the honest version. The interesting admission is that shadow AI is telling you your sanctioned stack isn't good enough. Right — the adoption spike is a diagnostic. It's employees voting on which corporate rollout failed them. If Tech Podcast Podcast is part of your Monday routine, consider subscribing and leaving a quick review wherever you're listening. It really helps other people find the show, and it helps us keep bringing you the briefings you count on.
If one of today's stories is still on your mind, you'll find links to all of them in the show notes, ready when you want to dig in a little further.
That's Tech Podcast Podcast for today. Thanks for listening, and we'll be back tomorrow. This is a Lantern Podcast.