
Formula for startup success: Find large highly fragmented industry w low NPS; vertically integrate a solution to simplify value product.
✓entrepreneur, investor, contrarian

Formula for startup success: Find large highly fragmented industry w low NPS; vertically integrate a solution to simplify value product.

What your VO2 max actually looks like: <15 ml/kg/min → Daily function and movement are limited. Difficulties getting dressed, standing up, and walking without assistance. 15–25 → Very limited fitness. Stairs are difficult, running is nearly impossible. 26–35 → Out of shape territory. You get winded carrying groceries or climbing stairs. Short jogs feel brutal. 36–45 → Functional, but still below true fitness. You can exercise, but endurance is limited. Cooper Test: ~2300m. 46–55 → Fit and healthy. Good energy, decent endurance, and fitness. Cooper Test: ~2700m. 56–65 → Extremely fit. The fittest person in a normal friend group. Running feels easy. Cooper Test: ~3200m. 66–75 → Elite amateur athlete territory. Most people can’t keep up with you. Fast recovery. 76+ → World-class athlete physiology. Most people can't comprehend your endurance. You can run laps around 99% of people.

🚨 BOOM: President Trump just announced a new QUAD SUPERINTELLIGENCE ACCELERATOR in Manila. He says the U.S., Australia, India and Japan are teaming up on superintelligence, and that America is LEADING THE WORLD by quite a bit. He's putting MORE THAN $85 MILLION into expanding trusted AI systems across the Indo-Pacific. Trump says breakthroughs that took 40 years are now coming in FIVE MINUTES, and he calls it bigger than the Industrial Revolution. He also announced a new Quad exercise featuring maritime patrol aircraft. The future of this technology belongs to free nations, not Beijing.

While everybody was watching the Nvidia drama, Trump's team quietly built something big. The US just signed the "San Francisco Declaration" with eight allies, Japan, Australia, Israel, the UK, South Korea, India, Qatar and the UAE, locking arms on AI and the technology that runs it. It's about cutting China out of the supply chains that matter most, the chips and the rare earths the whole AI race depends on. Germany, the Netherlands, and more are expected to sign on next week. This is the game under the game. Not a tweet, not a soundbite, a coalition. The kind of thing you only notice years later when it's the reason we won. Build the alliance now, win the century later.

Still the best analysis in all of tech (even though i often disagree w Ben re China policy).

La compañía de tech que más ventas consigue por empleado del mundo es....

For years, a widening trade deficit often reflected greater reliance on cheap foreign consumer goods, but the latest figures illuminate what many headlines obscure. The trailing 12-month goods deficit is 1.6% of GDP when excluding SI-related goods, its lowest reading in records dating back to 1994. When including SI-related goods, it stands at 3.5% of GDP, lower than any reading outside this current Administration since 2010. Recent import growth has been concentrated in capital goods that support investment and expand our capacity to produce at home, a shift increasingly reflected in stronger output and productivity across the American economy.

I have a major result to share. I’ve found conclusive forensic evidence that the top 4 Chinese AI labs—DeepSeek, Moonshot, MiniMax and Ant Group—have built their newest models on NVIDIA’s most advanced Blackwell chips, which are barred from export to China. This sounds insane, but I discovered that almost every AI model carries a hidden signature of the GPU that trained it. When a model is first set up, the starting values of its weights are laid out in a pattern that depends on the exact GPU. Each combination of GPU and seed number produces a unique pattern—a sort of fingerprint of the hardware used to create the model. Traces of that pattern survive training into the published model. Again, this evidence isn’t based on any insider information or leaking. It’s based on a forensic analysis of the model weights released by the companies themselves. I searched more than six trillion combinations of chip layout and seed number and matched them to open-weights models. In doing so, I found conclusive evidence that six Chinese AI models—including DeepSeek-V4-Pro, Kimi K3, and MiniMax-M3—were initialized on NVIDIA Blackwell systems. Four of the models match Grace Blackwell, the GB200 and GB300 parts in NVIDIA’s flagship liquid-cooled rack-scale systems. I then verified these results on real hardware. Running unmodified PyTorch with the seeds I’d recovered, a Blackwell B300 reproduced DeepSeek-V4-Flash’s untouched starting weights bit for bit, and a rented GB200 reproduced Kimi K3’s to a correlation of 0.999995. A Hopper H100 reproduced none of DeepSeek-V4-Flash’s, and changing the seed by one broke the match. (DeepSeek’s seed is 953049583 by the way.) The method also was verified to match 24 known open-weights models to the hardware that developers said they used. It is exceedingly unlikely that these results are a coincidence. Across six trillion combinations, the best accidental match should reach only about 8 sigma (i.e., eight standard deviations above what random chance would produce). The real matches range from 49 to 1,385 sigma, and even the weakest single test hits 20. These were data-center-scale training runs. With some analysis of release schedules and model details, I estimate that the largest, Kimi K3 and DeepSeek-V4-Pro, used 2,000 to 3,000 Blackwell GPUs each. Clusters of this size cost on the order of $150 to $200 million each to buy, and across these labs alone would add up to more than half a billion dollars of export-controlled GPUs, not counting experimental compute or otherwise. It’s quite likely that we see multi-billion dollar expenditure on export-controlled GPUs. I am highly confident in this result, and it has been independently verified by senior researchers at OpenAI and Anthropic. You can read the full report at unemploymentlabs.com/checkpo…. There is much more I’ve identified about these models through other forms of checkpoint forensics like this, and will have more to share soon.

I'm increasingly convinced that maintaining high aerobic fitness may be one of the most important things we can do for brain aging. Brain MRI research has linked higher aerobic fitness with less grey-matter shrinkage across the lifespan in a region involved in attention and working memory. Older age was associated with smaller gray matter volume there among less fit adults, but not more fit adults. Larger volume in that region was also linked to better cognitive scores. Long-term observational evidence has associated each 1-MET improvement in estimated fitness with a 16% lower risk of dementia. Separate exercise studies have reported increased glucose uptake in brain regions vulnerable to early Alzheimer’s-related changes after HIIT, and improved brain blood flow after aerobic training. VO2 max is a capacity we can measure and train, which is why I keep coming back to intervals like the Norwegian 4x4. Taken together, these findings make aerobic capacity hard to ignore.

America should not be pressuring @Israel to show restraint against Iran. We should be working to stop Iran's nuclear drive.

18% of NYC private employees work at "non-profits" The scale of fraud taxpayers like me are funding is truly unimaginable

.@rokhanna Americans are struggling while you are insider trading?

pretty amazing trajectory for the biz @ImprintPayments founding --> ~$2B val + on track for ~$400M GAAP rev in 6 yrs + feels like the type of biz that compounds at scale...

HUGE fintech round as @ImprintPayments raised $130M in a round co-led by @khoslaventures + @psumvc at a $1.88B val (up from $1.2B last Dec) Imprint builds + runs co-branded credit card programs for brands, owning the full stack from issuing + processing to rewards + loyalty some stats on the biz: > expects ~$400M in GAAP rev this year, up 150% (w/ headcount only up 30%) > 2M+ cardholders, up from ~15K 3 yrs ago > 130M+ txns processed to date customers include Kroger, Albertsons, Shell, H-E-B, @bookingcom, Rakuten + now Fanatics upstartsmedia.com/p/fintech-…

One month into my time at @factory and one core assumption is validated: The race to the enterprise will be won by AI companies that build the best infrastructure. To unpack a bit: -Enterprises are riddled by legacy, debt and off the shelf tooling. Integrating, modernizing and making these systems AI enabled is a key challenge. This isn’t a simple vibe coding task, it takes serious infra chops. -The infra is the foundation on which the product lives. If it is shaky, the customer experience will suffer and customers will churn. -Day 2 support really matters. You can’t just deploy and forget. Durability, reliability and support are key. -Standardizing the deployment configurations from the beginning is a must. If each customer cluster is a snowflake, you are doomed. This is why I’m pumped about @factory. We are solving the hard infra problems head on and not kicking the can down road. The team is stacked, come build with us 🚀

Another outstanding NYC company.