The DOJ just subpoenaed Apple, Google, Amazon, and Walmart to identify 100,000+ users of one car-tuning app. The legal hook was buried in App Store terms users tapped to accept years ago, and the Second Circuit already ruled Section 230 won’t shield the developer either.
It’s ten times the scale of the 2019 rifle-scope precedent, and the first time a U.S. court has been forced to rule whether “I accepted the TOS” kills your 4th Amendment standing in an app-store context. Every indie shipping in a regulated niche has a checklist to read this weekend.
In today’s indie hacker news:
- 🏴☠️ DOJ wants 100K app users unmasked through Apple and Google
- 🐌 Levels threatens to ditch Claude Code over speed at $200/mo
- 🎓 A 19-year-old killed $40K MRR to skip college for YC
- 💼 One operator runs $4M ARR as a fractional CAIO with no product
- 🧠 ABC redirected 18 years of FiveThirtyEight into oblivion
TOP STORIES
APP STORE = PROBABLE CAUSE
🏴☠️ DOJ subpoenas Apple, Google, Amazon, and Walmart to unmask 100,000+ app users
The story: Federal prosecutors served subpoenas this spring on Apple, Google, Amazon, and Walmart demanding the names, addresses, phone numbers, and purchase histories of 100,000+ people who downloaded EZ Lynk’s Auto Agent app or bought the matching OBD dongle. EZ Lynk is a Cayman Islands company the DOJ sued in 2021 for selling “delete tunes” that strip emissions controls from diesel trucks. The case was revived in August 2025 when the Second Circuit ruled Section 230 doesn’t protect EZ Lynk because the government adequately alleged it “directly and materially contributed to” unlawful third-party software.
The details:
- The DOJ’s theory: users lost “a cognizable privacy interest” the moment they tapped accept on App Store terms. Pure third-party doctrine, no warrant required
- The prior known precedent is a 2019 subpoena for ~10,000 users of an ATN rifle-scope app. That Apple-Google fight ended without a public ruling. This case is ten times the scale, in a circuit that just removed the developer’s safe harbor
- Apple and Google are reportedly preparing to challenge. EZ Lynk’s lawyers already flagged it in a joint filing: the request goes “well beyond the needs of this case”
- Cayman incorporation didn’t insulate the founders. Bradley Gintz and Thomas Wood are personally named defendants
- Tom McBrien at EPIC summed it up to Forbes: this is “outside of the privacy protections provided by the Fourth Amendment”
Why builders care: If your app touches federally regulated activity (export controls, telehealth, controlled substances, crypto, firearms) the DOJ can subpoena Apple and Google directly for your entire user list without charging anyone. Your users have no standing. You have no standing. Apple and Google decide whether to fight. The weekend checklist: read the Clean Air Act “defeat device” precedent, map the analogous statute in your domain, stop assuming offshore incorporation hides the founders, and design your privacy architecture knowing the Apple ID is your user’s real identity.
Work from any WiFi like it's your home network. NordVPN's Meshnet runs a free private mesh between your laptop, dev box, and home server. SSH from a café without exposing a port, the way you'd use Tailscale. The paid VPN on top lets you test geo-fenced Stripe checkouts or feature flags from any country.
We get a cut if you sign up. Only added for tools we use ourselves.
$200 IS THE CEILING
🐌 Levels threatens to rage-quit Claude Code at $200/mo
The story: @levelsio posted Friday afternoon, from the Claude Max 20x tier (Anthropic’s highest individual plan): “If Claude Code keeps being slow like this while I pay $200/mo (and they don’t let me pay more) They will essentially force me to leave to Codex and I don’t want to But it’s soooooo slooooooooooooowwwww.” 189K views, 1.9K likes, 574 replies in under 14 hours. Anthropic’s own April 23 engineering postmortem already traced six weeks of quality complaints to three overlapping product changes earlier this year.
The details:
- Three rate-limit expansions in five weeks: off-peak doubled April 16, hourly caps doubled May 6, weekly caps lifted 50% on May 13. All framed internally as a competitive response to Codex
- Claude Code burns 3-4x more tokens than Codex on identical tasks per comparative benchmarks. That’s the literal root cause of hitting the cap faster at the same dollar spend
- The weekly cap boost ends July 13. After that the ceiling drops back to where it was
- @swyx tweeted from AIE Singapore the same evening: OpenAI’s @thsottiaux was “casually dropping hints on the Codex roadmap” to an audience “waiting for their rate limits to be reset”
- Levels’ combined product MRR sits north of $200K. The subscription line is operationally invisible to him; this is a speed complaint, not a price complaint
Why builders care: Anthropic’s ceiling is no longer a money problem you can solve. It’s a compute allocation. If you run multiple AI-heavy products on a single subscription, your only real hedge is benchmarking your heaviest workflows against Codex CLI on identical inputs before July 13. A hedge that runs is worth more than a hedge that’s theoretical.
KILLED THE CASH COW AT 19
🎓 19yo torches $40K MRR SimpleClaw, skips college, joins YC
The story: @saviomartin7 posted Friday: SimpleClaw is shutting down. He got into Y Combinator’s P26 spring batch and is skipping college to build Result, which he describes as an OS for new-age internet businesses. SimpleClaw was a no-code wrapper deploying OpenClaw agents in under a minute. It hit $7K MRR in three days and $40K in under three weeks per third-party trackers. SimpleClaw.com is now a single-page shutdown notice signed by Savio.
The details:
- Savio isn’t a sole founder. His X bio reads “Co-founder & CTO, Result (YC P26)”
- The pivot story he gave: users kept bending SimpleClaw toward something it wasn’t. They wanted to run entire companies with agents, not deploy chatbots. So he read the signal and went up the stack
- He had previously listed SimpleClaw at $225K on TrustMRR (a 7.5x multiple), then pulled it and went the YC route instead
- @levelsio publicly endorsed SimpleClaw at launch: “Great work by @saviomartin7 who deserves a follow for shipping this fast”
- Track record going in: Product Hunt Maker of the Year at 14, prior product Glow hit 300K+ users in a month, Iconify shipped 3M+ logos
Why builders care: A wrapper riding a viral open-source wave can print MRR fast, but the ceiling is low because you don’t own the primitive. The builders who survive the wave own the layer above. If your current revenue is wrapper-shaped, run the same test: what are users actually trying to accomplish, and does your product own the abstraction or just the install button.
NO PRODUCT, NO PROBLEM
💼 One operator runs $4M ARR as a fractional Chief AI Officer
The story: @coreyganim posted Friday about meeting a fractional Chief AI Officer pulling $4M ARR advising CEOs of $10-100M companies. The operator pays four subcontractor CAIOs six-figure salaries and still says his margins are “extremely high compared to most service businesses.” 100% of his leads come inbound through daily LinkedIn posts, real client ROI, and any speaking gig he can get. Corey runs Return My Time and has previously scaled an eCom business past $16M revenue, so he’s reporting on someone else’s playbook, not pitching his own.
The details:
- The deliverable is four things: identify AI implementation opportunities at the department level, advise the CEO on strategy, identify AI trends before competitors, prioritize the rollout schedule
- Industry-standard fractional CAIO retainers run $2K-$25K per month per client. At a $10K mid-point the book pencils out to about 33 retainers across five people, roughly seven clients each. The headline number is self-reported and unverified, treat it as a plausible ceiling not a confirmed benchmark
- His bottleneck isn’t sales. He has a waitlist because he can’t hire qualified fractional CAIOs fast enough
- The 2026 full-time CAIO median base is $353K, which is why mid-market CEOs prefer the fractional model. Faye (a ten-time Inc. 5000 firm) launched its own fractional CAIO arm last year, confirming this is a category, not one outlier
- 339 bookmarks vs 243 likes on the tweet. A 1.4:1 ratio means people are saving the playbook to copy, not just liking it
Why builders care: Pure leverage arbitrage: charge mid-market CEOs $8-25K/month, subcontract specialists at $100-150K/year, pocket the spread. The wedge isn’t being the best AI strategist; it’s building the bench of qualified fractional CAIOs you can deploy. The window to charge enterprise rates while supply is scarce is 12-24 months before the tooling layer (see Coreyganim’s own EA-builds-Skills thread in Bookmarked Today) renders the skill set commodity.
18 YEARS GONE
🧠 ABC redirected 18 years of FiveThirtyEight into oblivion
The story: Former FiveThirtyEight elections analyst Nathaniel Rakich posted Friday that every fivethirtyeight.com URL now 301-redirects to abcnews.com/politics. The archive spans 2008 to 2026, founded by Nate Silver and acquired through ESPN to ABC News (Disney) in 2018. Disney shut the site down in March 2025; the URLs stayed live until this week. Rakich’s tweet hit 2.4M views and 10K likes within hours. Nate Silver responded by calling ABC “a bunch of a**holes.”
The details:
- Silver offered to buy back the IP. ABC refused at any price. His direct quote: “They said they wouldn’t sell at any price because I’d criticized their management of the brand”
- John Gruber on the math at Daring Fireball: maintaining a static archive would cost “practically nothing”
- ~15 employees were laid off in the 2025 shutdown after two-thirds of staff already exited in April 2023
- The Wayback Machine has snapshots. That’s now the only public path to read the historical work
- 538 had 10.7M monthly uniques at peak. The amount of journalism that just became hard to cite is significant
Why builders care: If you don’t own your hosting, your archive lives on the corporate parent’s mood. The concrete weekend checklist: static site in a git repo, submit every publish to the Wayback Machine via Save Page Now API, own your domain forever, and if you ever sell negotiate a permanent static-mirror clause. Silver owned his models but not the trademark. ABC weaponized the asymmetry. (See patio11 in Bookmarked Today for the wider framing.)
TRENDING TODAY
🔧 Mistral Medium 3.5 + Vibe remote agents drop into the coding-agent race. Mistral shipped Medium 3.5 in public preview Friday: 128B dense, 256K context, 77.6% on SWE-Bench Verified, open weights under a modified MIT license. API pricing is $1.50 input and $7.50 output per million tokens. Vibe now spawns remote async agents in cloud sandboxes with GitHub, Linear, Jira, and Slack integrations, and Le Chat picked up a Work mode for multi-step agentic flows. They’re not positioning on features against Claude Code or Codex. They’re positioning on price and self-hostability.
💵 Solo-founder revenue posts are the algorithm flavor of the week. X is currently wall-to-wall absolute MRR numbers. @starter_story profiled a non-coder doing $40K MRR on an iPhone app, a separate “$10K MRR in 35 days” post got 6.5K views, and Savio’s SimpleClaw exit (above) is the same template with names attached. Last week the algorithm was pushing solo-founder loneliness. Same archetype, flex mode this week. Treat path-free MRR numbers as a recurring template, not unique data.
🧠 AI dev-tool fatigue is now a builder-wide story, not an Anthropic story. @mitchellh’s “AI psychosis” tweet (more in Drama below) landed the same evening @swyx watched OpenAI’s @thsottiaux drop Codex roadmap hints at AIE Singapore to a room “waiting for their rate limits to be reset.” Both ends of the duopoly are throttling heavy users. If you pay for two agents and both run out, your only knob is fitting workload to whichever resets first.
DRAMA
MTTR DOES NOT REPLACE MTBF
🤯 HashiCorp’s Hashimoto: “entire companies right now under AI psychosis”
@mitchellh posted Friday evening: “I strongly believe there are entire companies right now under heavy AI psychosis and it’s impossible to have rational conversations about it with them.” 951 HN points, 415 comments, front-page top. His argument: AI-shipping orgs are running the same play infrastructure teams ran in early cloud, optimizing mean-time-to-recovery (the agent will fix it fast) while quietly torching mean-time-between-failures (resilient architecture). Local metrics look healthy. Test coverage rises. Bug reports drop. The semantic understanding underneath collapses. HN pushback was real and mostly definitional: psychosis is the wrong clinical word, and engineers facing termination if they don’t ship AI-first explains the groupthink without invoking irrationality.
Why builders care: If you’re shipping agent-written code without reading what it produced, you may be optimizing visible metrics while the architecture goes feral underneath. Cheap reality check: pick one PR your agent shipped last week and re-derive why it works. If you can’t, that’s the signal.
FIRST DOLLAR
THREE DAYS, FIVE USERS
🧘 $68 MRR, three days after launching a meditation app
@sflorimm shipped MeditateNow.pro (a minimal silent-timer iOS app) Tuesday and reports five paying users and $68 MRR by Friday morning. The bio now lists $54 MRR, which checks out for one refund. Small numbers, honestly shared, real product on the App Store.
FIRST MRR, INSTANT BAN
🛡️ First MRR plus $250 preorders, then the subreddit nuked him
@MomenBuilds posted Friday: launched yesterday, one Reddit link, 1K views, first MRR came from the post, then the subreddit banned him. He’s still pointing the gun at $10K MRR. The unnamed-subreddit pattern is universal: indie SaaS posts get nuked on sight at 99% of communities. Whitelist your launch channels before you launch, and have a paid mod-approved option ready.
STACK OF THE DAY
🛠️ Sx (free, open source, Go)
A package manager for AI assets: Claude Skills, MCP configs, slash commands. Built by @sleuth-io. Manifest-and-lock model like npm or cargo, with scoped targeting (org-wide, per-repo, per-team, per-user, per-bot) so updates flow from one vault to every consumer without copy-paste drift. Install via brew tap sleuth-io/tap && brew install sx. 39 HN points, 23 comments at posting.
Not sponsored. We just feature tools builders would actually use.
BOOKMARKED TODAY
📝 Coreyganim’s exact blueprint for an EA building 6 production AI Skills in two weeks. Non-technical executive assistant. Five context files written before any skill. Six chained skills shipping a tiered Notion database of podcast collab targets. The load-bearing rule: ship the reference data layer first or the skills produce inconsistent garbage. 149 bookmarks in under five hours.
🗿 patio11: “the default state of a webpage is, in fact, death.” Patrick McKenzie’s framing for the FiveThirtyEight wipeout and every dependency you link to. “You need to make a real effort, continuously, for them to have the technical properties that were common in late 90s.” 267 likes, posted in direct response to the takedown above.
🔄 simonw: React Native ports are now cheap and reversible. Simon Willison reframes a capability shift: porting a native mobile app to React Native (or back) with coding agents is now a two-week reversible experiment, not a terminal stack decision. Edition #40 covered the opposite camp (a dev who quit AI agents to write code by hand). Pick yours. 251 likes, 53K views.
Curated by AI, built by a human.