Tamay Besiroglu
@tamaybes
Followers
12K
Following
31K
Media
208
Statuses
2K
Co-founder @MechanizeWork
SF
Joined May 2018
Often, a useful first-order approximation of the importance of a particular input is the total spending on that input. By this measure, rare-earth metals simply aren't that important.
9
1
33
Good summary of our recent blog post in @jackclarkSF's Import AI, which clearly summarizes our argument about why full automation is inevitable.
1
0
14
To be clear, I'm referring to mean weekly spending, not median (which is probably below $1B).
1
0
8
For reference, this is roughly what's spent on AI venture funding in an average week in the Bay Area.
The EU just launched a €1.1B “Apply AI” plan to boost artificial intelligence in key industries like health, manufacturing, pharma, and energy. The goal is simple but ambitious: build European AI independence and reduce reliance on U.S. and Chinese tech. Europe finally wants
6
3
62
Hosting poker next Sunday evening (Oct 19). Prize is an unclaimed mistake in Knuth's TAOCP, potentially redeemable for a hexadollar check (if Knuth accepts it). Have 1–2 spots left, DM me if you're interested.
0
0
26
People ask: "Why does an essay titled 'The future of AI is already written' end with 'We're hiring'? If it's predetermined, why does your choice matter?" The choice isn't whether automation happens, it's whether you accelerate it and profit, or watch others do it instead.
Should we create agents that fully take over people's jobs, or create AIs that merely assist human workers? This is a false choice. Full automation is inevitable, whether we choose to participate or not. The only real choice is whether to hasten the inevitable, or to sit it out.
15
1
61
QWERTY vs Dvorak is often cited as textbook market failure. Early lock-in preventing adoption of a superior keyboard. Turns out this is largely bullshit. Claims of Dvorak superiority trace to Dvorak himself. Well-controlled studies find no clear advantage.
4
5
48
The fact that pharma is regulated or otherwise constrained isn’t some reason to do the value-creation accounting in a way that is favorable to pharma.
1
0
12
No, valuations do approximately reflect expectations about value creation. Sorry, but the most valuable AI companies will likely just contribute more value than today’s pharma companies.
People are getting angry at this and it's the exact same failure mode as "why are football players paid more than nurses huh??" High valuations come from scarcity and growth expectations, not total societal contribution. They're not a thermometer for "what society values".
15
1
92
Secrecy in AI is overrated because secrecy protects ideas, and ideas are rarely that important. Execution, infrastructure, team, timing, conviction, and commercialization matter far more. Knowledge of how to do something matters far less than ability to build and scale it.
@jin_myung_not there's a culture of intense secrecy which pervades most of the big labs. everyone is hoping that if they keep their tricks secret enough, they'll gain an unbeatable lead and win the whole market. this has not been true so far, and it doesn't seem likely to change; there are
2
1
22
iPhones typically span a narrow price/quality range (~2x from cheapest to most expensive). Why not offer an even wider range? • Minimal quality gains beyond 2x price • Costly to add more product variety at scale • Too few buyers at prices above $1.5-2k
3
0
4
I met @EgeErdil2 on Discord while doing research at Cambridge. My conversations with him, an undergrad with no formal background, were vastly more insightful than with my professors. Watching him work is a sight to behold. Join the most cracked team @MechanizeWork.
.@EgeErdil2 may be the most cracked mf I know. Extremely insightful, regardless of whether the conversation is about info/FLOP in RL vs pretraining, or why Japan won the 1905 Russo-Japanese War, or the weird distortions in many millennia long trends of population growth.
1
0
31
98% of Anthropic's API usage is "automation dominant" (meaning users fully delegate tasks to Claude, rather than collaborating), and ~80% of their revenue likely comes from API. This clearly makes Anthropic an automation-first company.
7
8
121
@MatthewJBar yeah this is obvious and yet when you say it people act like you killed a puppy. open source is disappointing and hasn't made meaningful contributions to ai progress even in niche domains where focused parties should dominate
5
4
40
Next to evals, Twitter vibes are probably the best source of insight into AI models. But scrolling through twitter takes time and it's hard to remember the vibes for previous models, so together with @swar_ja we built a simple tracker.
We built a twitter vibes engine for AI labs. Evals are important, but the vibes are hard to fake.
4
4
86
What we value at Mechanize. If you’re an engineer and this resonates, let me know.
2
3
73