Senior Solutions Software Engineer @ North American Tech (8 yrs). Translating Tech field insights to investing edge. š¢16% CAGR since portfolio inceptionš¢
In past 10 yrs, S&P500 biggest winners:
1) Nvidia (+23,671%)
2) AMD (+10,358%)
3) A??N (+3,005%)
4) Arista Networks (+2,750%)
5) Broadcom (+2,510%)
No. 3 breaks this pattern, touching neither AI GPU boom nor data center infra.
It also isn't part of the Mag7. This company is:
$AXON
To prepare for this ~8,000 word free deep-dive analysis available at https://t.co/xZyMEvcUzE, I went through more than 25hrs of video footage on the Boldly Go Podcast and almost every interview @AxonRick has ever done, as well as decades of SEC filings and news articles dating back to 2005.
With the stock down -42% from ATH, it is an interesting time to be studying this company.
Enjoy! š¢
Pat Dorsey (Dorsey Asset Management) on buying $AXON in Q3 2026:
"I've followed Axon for some time and have been impressed by its ability to complement hardware sales of tasers and body cams to law enforcement agencies with broader recurring software and services revenues from those same customers. That's made it a much more attractive business with deeper and stickier customer relationships...
Spending for public safety, despite the political rhetoric, tends to grow consistently at a mid-single-digit rate. And Axon should be able to grow much better than that because it offers a real value proposition to public-safety officers...
At a price that incorporated AI fears that we believe will prove to be overdone, we bought the stock in the third quarter of this year."
___
Source: Value Investor Insight, [Protecting the Castle October 31, 2026]
I have a major result to share.
Iāve found conclusive forensic evidence that the top 4 Chinese AI labsāDeepSeek, Moonshot, MiniMax and Ant Groupāhave built their newest models on NVIDIAās 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 patternāa sort of fingerprint of the hardware used to create the model. Traces of that pattern survive training into the published model.
Again, this evidence isnāt based on any insider information or leaking. Itās 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 modelsāincluding DeepSeek-V4-Pro, Kimi K3, and MiniMax-M3āwere initialized on NVIDIA Blackwell systems. Four of the models match Grace Blackwell, the GB200 and GB300 parts in NVIDIAās flagship liquid-cooled rack-scale systems.
I then verified these results on real hardware. Running unmodified PyTorch with the seeds Iād recovered, a Blackwell B300 reproduced DeepSeek-V4-Flashās untouched starting weights bit for bit, and a rented GB200 reproduced Kimi K3ās to a correlation of 0.999995. A Hopper H100 reproduced none of DeepSeek-V4-Flashās, and changing the seed by one broke the match. (DeepSeekās 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. Itās 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 https://t.co/YOtzND0Jqk.
There is much more Iāve identified about these models through other forms of checkpoint forensics like this, and will have more to share soon.
I believe the comparison of YouTube watch time to Netflix is misleading investors.
Consider a few examples: This "Chill lofi mix" is almost 2 hours long (length of a movie) has 35 *million* views in just four years. And this is one of hundreds of these types of videos.
People throw these up on their computer monitor as they're working, studying, working out, doing laundry. It sits in the background and adds some ambiance while they're actively working on something else.
In fact it's become a modern trend to throw up one of these videos on a TV in your room, living room, to set a 'vibe'. You could let it play for 8-10 hours a day, non stop, as background ambiance. (I know people personally who do this).
YouTube counts every hour of this as an engaged hour that is directly related and calculated the same as a Netflix viewer watching East Of Eden.
You realize the issue here, right? Well, it goes further than that. People use YouTube constantly as a substitute music-player. One of the most common use cases is to find a music playlist and play it right from youtube.
3rd party firms have estimated that 30-40% of YouTube's total watch time is from music alone. Netflix does not even compete in Music, yet this isn't broken out by YouTube because it paints a much better story to portray them as dominating the TV.
All this stuff like music and background videos contributes to their engagement numbers but it's fundamentally different than the type of content Netflix produces. This is why I believe many investors in Netflix are being misled.
We've made it easier to get your repo running in Claude Code cloud sessions on Pro, Max, and Ultra.
Claude figures out what dependencies, network access, and API keys it needs, then saves the environment for next time. API keys are saved securely, so Claude never sees them.