A quick intro since I haven’t been active on X or LinkedIn. I’m a neurologist. I built and sold Alertive to Carbon Health for multiple eight figures. Had two daughters along the way.

Don't worry team. I'm technical

Math checks out

Wow, DeepSeek is prepping an IPO. The number that should make you pause isn't the valuation. It's how little of the company the founder is giving up to get there which would be a dream for most VC-backed startups. Chinese AI lab DeepSeek is raising at a $71B pre-money valuation and targeting a mainland China IPO filing later this year, with a public debut penciled in for 2027. The valuation timeline is impressive: → Early 2025: ~$8B (right after the model that wiped ~$600B off Nvidia in a single day) → June 2026: first-ever external round, ~$7.4B raised at ~$50B post-money → July 2026: new talks at $71B pre-money — a 42% jump in roughly a month For three years, Liang Wenfeng funded DeepSeek entirely through his own hedge fund, High-Flyer. No VCs. No outside capital. When he finally opened the June round, he wrote ~$3B of it himself...I personally love to see this kind of alignment but it's definitely risk-on. Liang still owns somewhere between estimated 70-80% of DeepSeek after that making him the richest paper-billionaire in AI. DeepSeek building its own inference chip like the others foundation models. In summary, a hedge-fund founder built a frontier lab with zero outside money, matched US labs under chip export controls, engineered a cap table where no investor gets a vote, kept ~4/5ths of the equity — and is now vertically integrating into custom silicon. I'd speculate he's doing this to avoid a Jack Ma moment and get liquidity to shareholders and governance-control given China-US politics and the verbiage around 'winning war on AI' from this admin. Curious to know do you guys think? Why doesn't deepseek just stay private longer?

Only a few years ago this would’ve been a crypto ticker on his watch.

You know your in SF when the uber driver asks, “have you checked out Poe.ai?”

The cost of attention is going up. Ads are going to get more expensive every single year (a reason to be bullish meta) It's just supply and demand. As the cost to start a business, build, and ship an app or product goes down, the demand for attention is going up. Every industry gets more competitive over time. More companies show up. But the ad inventory stays roughly the same. So the question becomes: how do you lower your customer acquisition cost while everyone else's keeps climbing? You build brand. Brand is what makes people come find you. Someone searches a generic keyword in your category and thinks of you. Or better yet they search your exact name. That's demand you don't have to rent. Okay, so brand is worth building. Brands drive revenue in a few simple ways: Word of mouth Content marketing Referrals And only one of those can you control on day 1. Content marketing is the lever. The best of it sits right at the intersection of education and entertainment. Name the pain. Name the outcome. Paint the future your customer actually wants, then position your product as the bridge to get there. Make that content, and you build brand by default. Here's the part most people miss: you also manufacture word of mouth. The thing you "can't control" starts compounding because of the thing you can. I think about this as digital gravity. You're building digital mass on the internet. The more mass you have, the more gravity you pull in your category's attention and the more people start orbiting you. Go build your content engine. It's what I'm trying to do too 🫣. P.S. The flywheel of building getting customers > help them > referral to new customer takes time. As the world is shifting fast and we're trying to speed up b2b sales with brandjet.ai

Here goes that idea :)...

Alex’s team is one of the best!

More states should do this!

Funny how all the AI buble talk disappeared.

Anthropic just surpassed OpenAI in revenue for the first time. January 2025: $1B ARR >>April 2026: $30B ARR OpenAI is at $24B and missed reportedly missed sales targets. The company built by people who left OpenAI just lapped it which is a testament to focus and execution (accounting issues aside). Look at the developer layer. Claude Code launched mid-2025 with Cursor already near $1B ARR built mostly on top of Claude Code. Both Cursor and Anthropic have had negative gross margins. Cursor's problem is structural, they pay API costs they cannot escape hence the xAI deal. Anthropic's is reportedly gross margin positive now while being compute constrained. By April 2026 these products are neck and neck at $2.5B. Claude Code erased a multi-year gap in 12 months. But there's no rest for any of these teams. it’s not even close to over. Opus 4.7 is not as great as expected and codex is pulling ahead (see screenshot). Soon this subsidy will stop and prices will have to go up…for now it’s a lot like competing for the network effects just like lyft/uber. Companies building AI that are focused still have a huge opportunity if they can keep their cogs low. Using openrouter or together.ai might be a good step for now but ultimately tokens might cost more until datacenter demands catch up. Our team at Brandjet.ai's is doing this by building our own infra. The next 18 months will be really interesting!

Ian's on to something with Summation for sure. Couldn't be more excited for what he's building!