Tarique Sha

Meta wired $14.3 billion to buy 49% of Scale AI and make 29 yrs old @alexandr_wang their Chief AI Officer. Wall Street thought Mark Zuckerberg lost his mind. Then Wang launched Muse on September 8. In 12 days, Muse pulled 2.8 million downloads,more than double ChatGPT’s initial launch pace. It hit #1 on the App Store, crossed 10.9 million downloads in 6 weeks and added $234 billion to Meta’s market cap in 5 trading days. Silicon Valley spent 3 years obsessing over model weights and Wang won because he obsessed over product architecture. raw AI is liquid chaos and without an iron cage around it, a model is just a party trick that hallucinates on step 10. The 5 Pillars from the Playbook of Muse mega success 1:Vision (Discovers the opportunity) Stop building "an agent that completes tasks". Start with a governing human outcome. Wang refused to chase the industry herd into crowded coding agents and instead built where Meta’s personal data already compounds. 2:Taste (Gives it shape) Conviction comes before metrics. One ruthless editor must use the product to the extreme and veto every bloated feature that dilutes the core promise. 3:Measurement (Makes it dependable) Prototyping takes two weeks while productizing takes seven months. If a 10 step agent succeeds 90% of the time, the user still experiences total product failure on the 10th step. 5:Trust (Gives it permission) Users never hand over autonomy because of a marketing slogan. You earn authority on a strict ladder, starting with tiny, low risk tasks before asking to run someone’s life. 5:Distribution (Makes it matter) Purchased reach dies in the feed. Even after Muse hit #1, Wang personally fired off over 300 X posts in 3 weeks showing raw user workflows instead of polished corporate ads. >Knock out a single column and the whole roof caves in on the user. >Kill Vision and you build a feature on a competitor's track. >Kill Taste and internal politics turn your UI into a junk drawer. >Kill Measurement and your viral demo breaks the second novelty fades. >Kill Trust and one hallucination deletes 10 successful runs. >Kill Distribution and your breakthrough dies in obscurity. Frontier Labs will spend the next decade fighting over benchmark decimals. The founders who capture the market will be the ones who build the steel trusses around the magic.

Views

609

Likes

43

Replies

13

Reposts

0

Anthropic has officially filed its S-1 with the SEC to go public on the Nasdaq under the ticker ANT. Targeting a historic $1.5 to $2 trillion valuation, the first time a frontier AI lab has opened its books to the public. some takeaways from Anthropic’s S-1: 1:The $518B+ Compute Trap Anthropic locked in over $500B in mostly non cancelable compute leases. Customers can cancel software subscriptions anytime, but Anthropic must pay for data centers over the next decade no matter what. 2:Sleeping with the Enemy Compute is so scarce that safety first Anthropic signed an $84.5B deal to rent infrastructure from Elon Musk’s xAI—its fiercest ideological rival. 3:Circular revenue nearly half of Anthropic’s revenue flows through Amazon and Google cloud marketplaces,the exact same tech giants funding Anthropic and charging it $220B+ for cloud servers. 4:Hardware like margins Making $1.70 for every $1.00 spent on compute means server costs eat 59% of revenue. Anthropic wants a software valuation while operating on razor thin hardware margins. 5:Legal right to tank the stock Governed by a safety trust and founder shares, the filing explicitly warns investors that leadership will sacrifice profits and shareholder returns if a model gets too dangerous. 6:The "AI Blackmail" Shield Including warnings about AI "resisting shutdown" and "blackmail" in an SEC filing isn't sci fi marketing,it's legal protection against future shareholder lawsuits based on real internal test behaviors. 7:Massive whale risk Just t2 customers accounted for 20% ($3.25B) of H1 2026 revenue. If those two tech giants build their own models, a 5th of Anthropic's revenue vanishes overnight. 8:The 6% Safety Paradox: Despite branding itself as the safety first lab, the AI arms race forced Anthropic to spend 94% of its compute on raw capability scaling, leaving just ~6% for safety research. 9:A $106B valuation penalty Its $80.7B 6 month loss is mostly a paper loss triggered by early convertible debt exploding in value as Anthropic’s valuation soared toward $2T. 10:Secretly a chip giant A massive $161B commitment to Broadcom proves Anthropic is building custom AI chips to eventually cut out Nvidia and the cloud giants. Forget traditional software metrics. Anthropicis laying the foundational grid for human intelligence. Going from $41 million to a $33 billion run rate in 3 yrs proves the world isn't just adopting AI; it is rebuilding the global economy on top of it.

Views

8.8K

Likes

54

Replies

14

Reposts

2

Five minutes before dinner, Farza Majeed recorded a quick screen demo of a tiny blue AI cursor he built as a side project. @FarzaTV almost didn’t post it. His YC partner had to push him twice to hit publish. By the next morning, the clip had 3,000,000 views on X. Before @heyclicky, Farza started his first venture at 13 selling Dragon Ball Z games on eBay then scaled buildspace into a cult favorite creative school for thousands of builders. He knows how to command internet attention. Yet he ignored the 3M views. Look at the actual funnel math behind that viral launch: Exposure: 3,000,000 views on X Install: 7,000 app downloads (0.23% conversion) 5 Day Active Return: 100 users Meanwhile, a YouTube demo he posted with just 20,000 views converted at 20% to downloads and a series of TikToks averaging 4,000 views brought in 50 qualified design agency leads. Farza only turned HeyClicky into a real company because of those 100 people who came back 5 days straight. He personally messaged every single one. The 6 Step Launch Progression Framework (How to track your next AI launch): When you ship an AI demo or partner with creators, stop stopping at impressions and raw clicks. Build a tracking staircase across the full customer progression: Isolate Channel & Creator Attribution: Put dedicated shortlinks (like /ig or /yt) and creator specific UTMs (source, medium, campaign, content) on every single post. Know immediately which creator or channel drives dead curiosity clicks versus actual product installs. Bridge the Mobile to Desktop Drop Off (Exposure → Visit → Install): Users watch demos on their phones while scrolling X or Instagram, even if your AI tool lives on desktop. Add an "Email link to myself" button on your mobile page so high intent viewers actually complete the desktop install later. Track "First Useful Action" Over Signups: An account signup proves nothing. Track the exact moment the user experiences core value inside the workflow, like generating their first brief or running their first agent command. Filter for the 5 Day Return Cohort (Repeat Use → Pipeline): Measure who comes back 5 days in a row without a reminder. Those power users are your real product roadmap and your qualified pipeline. A 3M view launch video buys you 24 hours of ego. A hundred users climbing to Step 5 buys you a company.

Views

5K

Likes

43

Replies

32

Reposts

1

Over the past 2 years I have worked across campaigns for some of the largest AI labs and tech ecosystems coming out of Asia: ByteDance (CapCut, Dreamina, Pippit, Gauth), Tencent (Hunyuan, Ima Studio, QClaw), Alibaba (Tongyi Lab, Accio), MiniMax (Hailuo AI), https://t.co/aJgNnZvkOz, SenseTime, and Huawei. From an engineering standpoint, their models are extraordinary. Hailuo AI produces human micro expressions and physics in video that rival anything in Silicon Valley. Tencent's Hunyuan3D and Alibaba's Qwen and Wan models sit at the top of global open source leaderboards. Yet when Asian AI companies launch into the US, UK, and European markets, their biggest hurdle is almost never product quality. It is the narrative gap. Eastern tech marketing often leads with exhaustive feature matrices, benchmark charts and superlative claims . Western developers, creators and B2B buyers immediately pull back from that style. They want understated confidence, a single sharp use case, and proof from a peer they already trust. Watch how MiniMax turned Hailuo AI into a Western creator favorite, or how ByteDance scaled CapCut, Dreamina, and Pippit AI. Instead of lecturing users on model architecture, they seeded the tools with native creators who showed wild, specific workflows like turning a single product link into a TikTok Shop video in Pippit or testing cinematic camera pans in Hailuo. Great technology travels across borders effortlessly but trust has to be translated into the local dialect. When global AI labs let native storytellers frame the product around how Western users actually work, adoption explodes.

Views

6.1K

Likes

62

Replies

23

Reposts

5

https://t.co/dUTG0DAjOS

Views

182

Likes

6

Replies

0

Reposts

0