The threat isnt that AI makes bad decisions. Its that the workflows AI is being dropped into were already wrong. Enterprises are automating the wrong thing at scale.
@wquist
English
Will Quist
8K posts
Will Quist retweetledi
We now have 30+ companies competing in the data centre CPU & TPU markets. All at astronomical multiples & valuations. Barriers to entry seem non-existent.
We have a 3-player oligopoly in HBM, with no new entrants even on the horizon. Barriers to entry are astronomical.
Hmm…
English
Will Microsoft offer 30% of their market cap to acquire OpenAI?
JJ@JosephJacks_
Anthropic is to OpenAI what Google was to Yahoo.
English
@Finding_Moats But we can agree that is always the smartest place to start right?
English
Not exactly my point though. What I was trying to say is that every time a stock drops, people suddenly remember valuation mattered — as if it was obvious all along. You understand the business first, then you worry about the numbers. Using a ratio on accounting earnings as the main filter for cheap vs expensive is, at least the way I see it, where things get a bit too simple. Same goes for DCFs on businesses with high barriers to entry and assets that are nearly impossible to replicate — the spreadsheet will never fully capture what makes those businesses special.
English
The problem — at least the way I see it — is reducing investing, with all its complexity, to a mathematical formula, however sophisticated it may be. How do you put in a DCF that $AMZN selling books would end up building AWS? That $DHR would go from an industrial asset play to a pure diagnostics and life science business? That $NFLX would disrupt Blockbuster and build a dominant global streaming service? That $TSLA would go from selling cars to selling autonomous humanoid robots? That $GOOG would acquire YouTube and turn it into one of the most profitable assets on the planet? The hard part is everything the formula can't see, IMO.
Gerhard Visagie 🇿🇦🇷🇼@Gerhard_Paladin
Best way to avoid overvaluation is to do a DCF valuation and IRR calculation. Even with simple assumptions .
English
Someone is gonna crush it with whatever the QA framework is for this era
Garrett Lord@GarrettLord
with a coding copilot, you see every output. you are the eval. with an agent doing compliance or underwriting or tax advisory, nobody is looking. talked to a dozen Fortune 100 companies this month. same story everywhere. they deployed agents. they have no way to measure if the agents are doing a good job. one told me the reason they think their AI works is because they've never tested it.
English
Will Quist retweetledi
more detail on why I think the AI labs are probably overglazed:
- I believe we are over 100% penetrated of the first slug of AI users. More will come with fine but hearing some dumb tokenmaxxing stuff. People are spending 2x their salaries (often by mistake) and not generating topline. Things like continuous web scraping, etc. are wildly expensive. This sets up for “AI optimizations” in the same way we saw “cloud optimizations” with $SNOW and $DDOG in 2022. AI is real but will take time to get the second mover cohort on board and the first movers are clearly “overusing”
- these labs are a boom for VCs who had mega funds they needed to deploy but no co’s who could take huge checks and an inexplicable allergy to hardware. along come mega capital intensive co’s that you can plow huge checks into bc you know the idiot next door will do it next round. these will be good IPOs but the price charts will look like $SNOW. the private markets extracted too much value too fast
English
Will Quist retweetledi
Ai is expensive bc it’s made by fancy people in fancy places and subsidized by fancy VC’s.
This is not priced in
Xiaoyin Qu@quxiaoyin
Why is @deepseek_ai 100x cheaper than @AnthropicAI? China is vertically integrated to be cheap. → Cheap model: token-optimized, aggressive caching, less GPU per query → Cheap chips: Huawei silicon, no Nvidia tax → Cheap energy: subsidized power, state-scale grid → Cheap talent: top researchers at a fraction of US salaries → Cheap economics: DeepSeek funded by trading profits, inference doesn’t need to make money The only thing they lag is performance. But as they get “good enough”, being cheap matters a lot, and being frontier keeps getting harder.
English
Amazing way to say 'we are gonna need to even MORE in training before we REALLY monetize it'
Sam Altman@sama
i keep thinking i want the models to be cheaper/faster more than i want them to be smarter but it seems that just being smarter is still the most important thing
English
Packy, you arent supposed to say that.
AND, it doesnt matter because... it doesnt matter.
Packy McCormick@packyM
If the rumored Anthropic at $900 billion round goes through, Anthropic will be the first company in history to cross the $900 billion mark with negative Free Cash Flow. Which is just an interesting data point is all.
English
Will Quist retweetledi
Anthropic will have the mother of all decelerations. They simply can't grow 50% month over month for long as they didn't secure enough compute.
MetaCritic Capital@MetacriticCap
$NFLX rallying 6% while groing 13% and trading ay 30x is a great example of @IdleLightkeeper evergreen tweet
English
Will Quist retweetledi
The cyber security oligopoly is dead.
AI changes everything:
• Attack volume explodes (automation)
• Attack sophistication leaps (reasoning)
• The surface moves (models, data, agents)
The current players are defending a world that no longer exists.
English
Will Quist retweetledi
Most cybersecurity AI takes miss the point entirely. They assume agents just make current workflows faster / smarter / stronger.
I can't help but believe in an autonomous world, the workflows themselves don't make sense anymore.
English
Will Quist retweetledi
Optimistic take: US healthcare is actually set up very well for the AI era.
GP shortage.
Best therapies and proceduralists in the world for actual interventions.
And a ton of admin bloat / bad incentives that create surface area to arbitrage.
English
English
Will Quist retweetledi
Slow Ventures' Slop Con NYC
luma.com/8boyfmmu
@slow @Jack_Raines @kevincolleran
@lessin
@wquist
@yrechtman
English
Will Quist retweetledi
The entire US economy right now is 7 companies sending a trillion fake dollars back and forth to each other
English