Editorial intro

The news cycle kept circling two themes today: big AI claims that beg for verification, and platform moves meant to rebuild user trust. We pick apart a jaw‑dropping revenue estimate for Anthropic, then look at Reddit’s heated debate over biometric checks and Microsoft’s quieter course correction on desktop AI.

In Brief

(No high‑quality, unambiguously verified scoop met our threshold today. Below are notable stories worth watching, but treat claims and reactions cautiously.)

Microsoft rolls back some of its Copilot AI bloat on Windows

Microsoft said it will remove some Copilot integrations from Windows apps and be “more intentional” about where AI appears. Copilot — Microsoft’s assistant powered by large language models — had been embedded across Photos, Widgets, Notepad and the Snipping Tool. The company framed this as focusing on experiences that are “genuinely useful,” while also promising better control for users and some performance fixes. Read the TechCrunch coverage.

Why it matters: Windows reaches billions. Pulling back signals Microsoft heard privacy and bloat complaints. For users, this means fewer surprise AI prompts and slightly more predictable behavior. For enterprises, it’s a reminder large vendors will iterate public AI rollouts more conservatively after backlash.

Slay the Spire 2: 9,000 negative reviews before a nerf is live

Developer Mega Crit pushed a beta balance patch for Slay the Spire 2, and players left over 9,000 negative Steam reviews in a day over a card change. The outrage centers on the Silent class card “Prepared” being made costlier. Complicating the response: players in China, who lack full Steam community features, may use review-bombing as the only visible outlet. Read more at PC Gamer.

Why it matters: It’s an example of how platform limits shape user feedback. For developers, balancing patch transparency with global community differences is now part of release planning.

Deep Dive

(These picks have big implications but rest on uncertain or contested information. Read critically.)

We're not paying enough attention to Anthropic adding $6 billion ARR in February

ARR — annual recurring revenue — is a way to annualize subscription or usage revenue. It gives a quick sense of predictable, repeating income.

A Reddit post made a startling claim: Anthropic, the private AI lab behind Claude, may have added roughly $6 billion in ARR in a single month. The original estimate is explicitly rough “napkin math,” and it’s based on private contracts, not audited filings. Still, if the order‑of‑magnitude claim were even half right, it would change how we think about enterprise AI adoption.

What was reported and why it matters

The figure comes from extrapolating known enterprise deals and pricing to estimate monthly usage that was then annualized. Industry outlets have also noted Anthropic’s strong enterprise traction; Axios wrote that Anthropic is "capturing over 73% of all spending among companies buying AI tools for the first time." But both Axios’s share figure and the Reddit estimate rely on private contract data and vendor reporting, not public statements.

Why you should be skeptical

  • Private ARR math can hide big assumptions about customer tenure, discounts, and usage spikiness.
  • Inference — running the model to produce answers — costs money. Inference cost here means the cloud and GPU bills necessary to respond to every API call. If Anthropic’s pricing or customer usage leads to high inference costs, revenue can be less profitable than the headline ARR suggests.
  • Big enterprise deals can be noisy: initial consumption can spike on pilots and later taper off.

What it would change if true

A genuine multi‑billion‑dollar ARR gain in a month would show monetization is outpacing a costly model‑training era. That would boost demand for cloud GPUs and network capacity. It would also sharpen policy scrutiny: the Department of Defense has described Anthropic as an “unacceptable risk to national security,” a designation Anthropic is contesting in court. That legal backdrop can affect contracts, partners and cloud hosting choices.

A practical analogy: think of ARR like estimating how much a café will make in a year after one busy month. A packed week in February doesn’t guarantee steady customers for twelve months. The cautious read: this headline is worth watching, but treat the $6 billion figure as a directional signal, not settled fact. See the Reddit thread for community reactions.

"really napkin math" — a top commenter, urging caution on the revenue extrapolation.

Reddit is weighing identity verification methods to combat its bot problem

Biometric verification — checking a person’s identity using biological traits like a fingerprint or face scan — is back at the center of a big platform debate.

Reddit CEO Steve Huffman floated several verification ideas on a podcast, including “lightweight” device checks such as Face ID or Touch ID and heavier options like third‑party validators. Face ID and Touch ID are phone features that verify whether the device’s owner is present; on Apple devices, those checks typically happen locally in secure hardware and do not send a face image to the company.

Why Reddit is considering this

Automated accounts and AI agents flood conversation threads, often to manipulate discussions or spread misinformation. Reddit wants to keep the site anonymous but also to ensure posts are created by humans. Huffman framed the tension plainly: the company wants to know “you’re a person” while preserving user anonymity.

The tradeoffs

  • Privacy vs. authenticity: Device biometrics can confirm a human is behind an account. But many users worry about giving platforms any kind of biometric data. Even if verification is local to a device, user trust matters.
  • Centralization vs. decentralization: Third‑party validators or decentralized identity systems could prove humanness without Reddit holding ID data. But they add complexity and new trust assumptions.
  • User flight risk: On Reddit, many long‑time users value anonymity. Push too far and you risk losing communities. One top commenter said Reddit is “my last bastion of social media,” capturing a sentiment shared by many.

What this means in practice

If Reddit adopts a device‑based check, expect a system that proves a human used a device at account creation or login, without necessarily storing a name. Think of it like a nightclub bouncer checking that you’re an adult but not writing down your home address. That approach can stop cheap bot farms while keeping usernames anonymous. But it won’t stop all abuse — determined actors can still route through multiple devices or exploit human click farms.

Key takeaway: Reddit is balancing two hard problems — stopping bots and preserving anonymity. Any path forward will test how much privacy users will trade for cleaner conversations. Read the Engadget writeup for the CEO’s comments.

"The most lightweight way is with something like Face ID or Touch ID... Part of our promise for our users is we don't know your name but we do want to know you're a person." — Steve Huffman

Closing thought

Big numbers and bold proposals make headlines. The harder work happens when companies translate claims into verifiable accounting, products into trusted systems, and platform promises into user‑friendly policies. Today’s feed is a reminder to ask two simple questions: What’s the evidence? And who pays the cost — in dollars, privacy, or trust — if the headline turns out to be hype?