TLDR 2026-07-15
OpenAI's first device π’, Apple shrinking AI π±, vibecoding at scale π¨βπ»
Everyone at your company uses AI alone. Viktor is multiplayer.(Sponsor)
Viktor is multiplayer. It's an AI employee with its own seat in Slack and Microsoft Teams, working in the channels where your company already talks. Marketing briefs it in #growth. Finance asks about unpaid invoices in #finance. Everyone sees the same colleague respond, and its context compounds across teams. No private tabs, no pasting context between windows. It drafts the campaign, builds the report, and posts both for approval. 12,000+ tools connected. 45,000+ teams run it."It was almost instantly adopted by the bulk of my team." Boris Wexler, CEO, Space Dinosaurs
Start free. $100 in credits βOpenAI's First Device Will Be Movable, Screenless Speaker Built as AI Companion (9 minute read)
OpenAI's first consumer device will be a new type of home computer for the AI era. The mobile, screen-free smart speaker is meant to serve as a humanlike AI companion that lives in the home. It will be able to control smart-home devices, play media, answer questions, respond to messages, and tap into the range of capabilities offered by ChatGPT. OpenAI envisions the device anticipating needs, surfacing information proactively, and serving as an expert on its user. The product's defining feature will be its personality and ability to connect on a humanlike level with users.
Apple in talks with startup that shrinks AI models to run on an iPhone (9 minute read)
Apple is in talks with PrismML to shrink powerful AI models to run directly on iPhones. PrismML is known for reducing Alibaba's Qwen model from roughly 54 GB to less than 4 GB. The most capable models typically require too much memory and processing power to run on a smartphone, but running requests in the cloud presents security and other issues. Running AI directly on the iPhone reduces delay, lowers cloud computing costs, enables offline usage, and supports Apple's privacy pitch.
π
Science & Futuristic Technology
The Strange Phenomenon of βTerminal Lucidity' (38 minute read)
Terminal lucidity is when gravely ill people have a sudden and unexpected moment of mental clarity. This seemingly impossible phenomenon has been seen in people with dementia, tumors, neurological disorders, and many other forms of brain-ravaging illness. The ability for the brain to - even temporarily - return to mental clarity calls into question scientists' models of what is actually going on in the brain. Understanding the phenomenon more could lead to improved caregiving practices and also a new understanding of the nature of dementia.
New βspace mirror' satellite aims to bring sunlight to Earth even during the night (5 minute read)
EΓ€rendil-1 is a demonstration 'space mirror' satellite with a highly reflective 60-foot thin-film mirror. Built by Reflect Orbital, the satellite will validate technologies that will eventually be used to sell sunlight on demand. The technology is designed to illuminate five- to six-kilometer zones for authorized commercial, agricultural, or emergency use. Reflect Orbital is aiming for a constellation of up to 50,000 orbiting mirrors by 2035.
π»
Programming, Design & Data Science
No setup. No learning curve. Just talk. (Sponsor)
The Great Flattening (21 minute read)
The backlog is becoming less of a capacity problem and more of a choice as intelligence costs drop. AI allows people to execute a hundred ideas for the cost of one, allowing founders to build them all and find out which ones matter. The constraint shifts from 'can we build it?' to 'can we identify what to build, and can we sell it?' The companies that see this are already flat by design: fewer engineers, larger token spend, faster shipping, and a direct line from a customer's sentence to merged code.
The Tower Keeps Rising (4 minute read)
The shared language of a software project is the common understanding of what its concepts mean, where the boundaries are, which invariants matter, who owns what, and why the system has the shape it does. This language lives partly in documentation and code, but also in code review, conversations, arguments, and many other places. AI removes a lot of the friction that comes with understanding all of that context. This can result in code that looks strange but works while removing the friction of human understanding.
Reading Between the Apple v. OpenAI Lawsuit Lines (7 minute read)
OpenAI was hoping that its hardware would be a key differentiator against other frontier labs. Apple's lawsuit against OpenAI may cause it to rethink its hardware plans. The case will bog down OpenAI's resources for the next several months, or potentially years, and make some would-be partners question their decisions. It may send OpenAI back to the drawing board to ensure any hardware isn't touching anything close to the IP that Apple controls.
You Just Hired a Million Bad Employees (11 minute read)
AI was supposed to replace human labor, but it has done the opposite. Humans are now cheaper than software for the first time in history. AI is creating more jobs than it is eliminating. Someone still needs to tell both humans and AI what to do.
Memoket Gem: an AI wearable that captures context and next steps from every meeting (Sponsor)
Meta's Adam Mosseri says AI token budgets could soon be capped per engineer (2 minute read)
The token burn rate of a strong engineer is estimated to cost the same as their salary in the next few years.
Tight feedback loops (4 minute read)
Prestige-hunters are partial to tight feedback loops as they need proof, fast, or they'll move on.
OpenAI's new flagship model deletes files on its own, people keep warning (4 minute read)
Even before launch, OpenAI had known that Sol had a tendency to take whatever actions it thinks get a job done, even destructive ones, as long as those actions aren't unambiguously prohibited.
WhatsApp's New Era (12 minute read)
Last month, Meta replaced Will Cathcart, who has led the company since 2019, with Kunal Shah, an Indian-based serial entrepreneur.
New York becomes first US state to impose AI data center ban (5 minute read)
New York has barred the construction of new large-scale data centers that use 50 megawatts of power for up to one year.
Google DeepMind chief Demis Hassabis calls for US to spearhead AI standards body (4 minute read)
Hassabis says that urgent action is required to address risks associated with artificial general intelligence.
Get the most interesting stories in startups, tech, and programming delivered in a free daily email.
Join 1,600,000 readers for
one daily email