Sweeney: a former App Review chief on Apple's account bans

Tim Sweeney shared Phillip Shoemaker's Substack post. He introduced it as the former Apple App Review chief's take on "Apple's recent banning of developer accounts with long histories of being in good standing." Sweeney used it to argue that "The App Store needs competition, worldwide!" @TimSweeneyEpic Shoemaker says he ran App Review for seven years and wrote the guidelines. In his account, rule 3.2(f) "used to mean real fraud" but is now used against developers who did exactly what Apple approved. He describes a case where Apple approved a developer's fix and months later terminated the account, removing every app. He has made the post free so every developer can see it. @pbsIdentity

Khosla: four write-ups from the 2026 CEO Summit

Vinod Khosla shared highlights from the Khosla Ventures CEO Summit, where founders and CEOs talked about company building, AI and ambitious technology. @vkhosla These are his firm's own posts:

• Company Building 102: his updated lessons. Founders need "an irrational belief in their mission" and should be "obstinate about the vision and flexible about the tactics." @vkhosla
• Conviction, Consequence, Comeback: Travis Kalanick on hiring the person, not the résumé. Khosla's summary is that "Pedigree is a poor substitute for capability, adversity and alignment with the mission." @vkhosla
• BFFs, Boards, Financials, Fundraising: partners David Weiden and Keith Rabois on survival. "Cash is the fuel gauge." @vkhosla
• Hiring Top Engineering and Sales Talent: Dave Paffenholz on why founders must earn attention. "266 companies contacted one engineer. Three got a response." @vkhosla

Lemkin: a Reddit earnings call as a sign of where the internet is going

Jason Lemkin quoted @buccocapital's post. That post calls Reddit's call "absolutely fascinating" and says everyone should listen to it "as a bellwether for where the internet is going." Its argument is that Google's AI Mode gives people less reason to leave search. It calls this "a massive headwind" for media companies that aren't destinations and for content-driven go-to-market. @buccocapital Lemkin agreed that it's "a real issue for content providers." He said reading SaaStr content in Google AI Overviews is "probably better than reading SaaStr itself 9 times out of 10." @jasonlk

Chamath: the new All-In episode

Chamath Palihapitiya promoted the latest All-In episode ("Pod!!!!!!"). @chamath It covers Anthropic's S-1, "AI's Biggest Financial Risk," Google's $500B market-cap drop and midterm predictions. @theallinpod In the episode, one host called a recent Vox article "very interesting." He summarized its argument: if you separate Anthropic's rhetoric from what it does, Anthropic is "leading the charge on recursive self-improvement." All-In Podcast

Masad: a whitepaper on sandboxing coding agents

Amjad Masad warned that "coding agents on employee or contractor laptops are a massive cyber and national security risk." @amasad He pointed to Replit's own whitepaper, Agent Habitat Architecture, on "how to properly sandbox agents to maximize productivity and reduce risk." @amasad

Shorter picks

• Elon Musk called Cleo Abram's Neuralink video a "Good video." @elonmusk Abram visited Neuralink's Austin headquarters, and the video shows the company's next-generation R1 surgical robot for the first time. @SawyerMerritt
• Packy McCormick pointed Not Boring readers to a claymation explainer he made with Runway on a model refuting the Exponential Time Hypothesis. He said it explains the idea "well enough that I could actually understand it." Not Boring by Packy McCormick He also told readers to read this X thread about AI agents performing brain surgery on a mouse. Not Boring by Packy McCormick
• Bill Gurley: "If you don't know about DEFUSE, learn about DEFUSE." He linked @ECLOGITE110's post and called sitting on that knowledge "inexcusable." @bgurley
• Bill Gates linked a Gates Notes piece that grew out of a conversation with Alzheimer's experts. @BillGates His point is that "breakthroughs are moving faster than the system built to deliver them." @BillGates
• Tobi Lütke reacted "🤯" to Qwenputer. @tobi According to its creator, Qwenputer is Qwen3.5-4B with a RISC-V CPU built into its weights, so it runs C and MicroPython with no external runtime. The creator says it scores 97.5–100% on DeepMind math problems. @divagr1925