In mid-2024, Pocket FM was at $200M ARR, growing 50% YoY, and we decided to kill its business model. Revenue flatlined for six months, we turned cash flow negative, and investors were writing us off. But then, Pocket's ecosystem exploded. Our reasoning and why this happened: What constrains Netflix's growth is content. They spend $17B+ on content production yearly, but many shows don't make it to the final stage. Of those that do, only a few become blockbusters. And those few are the reason subscribers stay on Netflix. So we started thinking. Pocket was growing nicely. However, what limited us was how many new users we could add and retain at scale. The only solution we came up with was increasing the number of blockbusters. Blockbusters drive user growth and retention at the same time. To get more blockbusters, Pocket FM needed more production. That's why we pivoted to only AI Content production. This did a number of things: 1. More content gets produced because we removed a lot of friction from doing so. It's easier for writers to maintain daily production. And for writers who still haven't found their hit show, testing and pivoting becomes less time-consuming. From taking several months to test even one show, you can now test 3-4 shows in one month. 2. Content quality increases. The more content gets produced and consumed on Pocket, the better the engine gets. We continuously monitor retention and other engagement metrics. When we detect something is working, our writing copilot learns it. 3. Unlocked hundreds of thousands of writers; Pocket's writing copilot is built to help writers without storytelling training. This is in contrast to Netflix's model. While their budget gets allocated among a limited number of studios, Pocket's supply side is getting flooded with writers. Over 500k writers are producing on our platform. 4. Made it easier to cross-create ads to attract users. Better shows are easier to sell to people who aren't yet using Pocket. And the same data that feeds our copilot goes on to feed the engine we use to create ad videos, which is also self-reinforcing. _____________________________ After making the change, ARR flatlined for six months and we started losing money. It was extremely painful, but then growth and the whole ecosystem exploded. - Hours of content produced per year went from 35k to ~2.5M. - New titles added yearly more than tripled from 120k to 400k+ - Millions of new users joined Pocket. The obvious risk was opening the gates to AI slop. That's why we have an extremely rigorous feedback and moderation process Ultimately, the only thing that matters is whether users are enjoying the content that's being uploaded. Metrics don't lie: - 12-month revenue retention went from 44% to 76%. - Avg daily streaming time went from ~25mins to 150+ mins. - DAUs/WAUs from 30% to 51%. I think we're extremely early, and I'm excited to see how Pocket's flywheels continue to accelerate from here.

From the Financial Times Anthropic July revenue: $65B annualized. Investors had modeled over $80B. OpenAI: over $40B annualized, up 35% quarter to date. GPT 5.6 reaccelerated them after a slow start to 2026. FT says OpenAI is gaining ground on Anthropic. Spending on Fable 5, Anthropic's most expensive model, has flatlined at about 11% of total Anthropic spend, two and a half months after launch. Opus 5, Anthropic's cheaper model, launched in late July and has already passed Fable 5 in business spending. $NVDA