Anthony Bardaro

50.3K posts

Anthony Bardaro banner
Anthony Bardaro

Anthony Bardaro

@AnthPB

PM² focused on capital markets and tech/media/entrepreneurship – leave your mark 👉 https://t.co/zlM0K4rIHR #dyodd #nia

Boston, MA Katılım Şubat 2015
625 Takip Edilen3.5K Takipçiler
Matt Slotnick
Matt Slotnick@matt_slotnick·
if inference ends up like electricity, what does that mean for products that make use of it? your lamp doesn't monetize on electricity consumption
English
9
1
30
4.9K
Anthony Bardaro
Anthony Bardaro@AnthPB·
Also Cloudflare: "cars are endangering horses" . . . "autonomous vehicles are endangering cars" And who can forget: "the internet is endangering newspapers" . . . "apps are endangering the internet" $net
Techmeme@Techmeme

Google's AI Search is endangering the open web; Cloudflare: between June 2025 and April 2026, human traffic to sites of businesses in many industries fell ~40% (@kateconger / New York Times) (Visit Techmeme dot com for the link and full context!)

English
1
0
1
326
Anthony Bardaro
Anthony Bardaro@AnthPB·
…wild how much uncertainty and how many variables stem from the Mythos nerfing; export ban/controls/suspension; and federal regulations¹ (EO 14409) thereof – widens the aperture of outcomes, e.g.: 1️⃣ Chinese (Kimi K3) and open models (Nemotron²/Inkling³) have closed the gap to US GA jagged edge (nerfed Claude Fable 5) to 8½~9 months⁴, but even tho intelligence leap from Fable to Mythos isn't huge, how far ahead of these GA models is Mythos at the real cutting edge? 2️⃣ what are supply/demand curves for that real cutting edge (Mythos)? 3️⃣ to what extent is Mythos' nerfing an intentional "natural break" or capacity cope (Anthropic's/USG's/etc)? 4️⃣ do other US labs have even better models that they're sitting on due to regulation, safety, competition, capacity, etc? 5️⃣ what are equilibrium efficiency; cost⁵ (TCO from training/capex/inference input > reasoning > output); and economies/diseconomies of scale for full stack infrastructure integration (Google/OpenAI) vs increasingly heterogenous (Anthropic) vs distributed edge computing (Qwen/Kimi/OpenClaw)? __ ¹x.com/i/status/20658… ²x.com/i/status/20339… ³x.com/i/status/20775… ⁴x.com/i/status/20788… ⁵x.com/i/status/20788… #datacenter #gpu #asics
Jukan@jukan05

After reading @deanwball’s piece, I had the opportunity to read several expert-call transcripts. Having done so, I concluded that his argument is, to some extent, mistaken. Here is why. Take DeepSeek as an example. Even though DeepSeek has open-sourced its model weights and parts of its software architecture, competitors would still find it difficult to replicate its cost advantage. That is because, while DeepSeek has disclosed most of its model architecture, the critical implementation details and operational know-how remain proprietary. As a result, even if Chinese or U.S. hyperscalers deploy DeepSeek’s open-source models on identical hardware, DeepSeek’s own deployment environment can achieve—and is already achieving—greater operational efficiency and a lower average inference cost. This efficiency advantage allows DeepSeek to price tokens through its official API below third-party platforms while still maintaining an API margin of 70%. Ultimately, China’s decision to release model weights cannot simply be characterized as dumping. Chinese companies may lack sufficient compute capacity to serve all the inference demand themselves, but they are not selling at a loss or failing to recoup their training costs.

English
0
1
1
167
Anthony Bardaro retweetledi
Anthony Bardaro
Anthony Bardaro@AnthPB·
@TheStalwart Because those aren't the only ingredients¹ and they have access to those chips²... ¹x.com/i/status/20451… ²x.com/i/status/20575… #kimi
Anthony Bardaro@AnthPB

…atm, US's half pregnant "China Chip Ban" is a goldilocks situation for the CCP: 1️⃣ sufficient supplies of American chips getting smuggled in¹ 2️⃣ nearby/offshore datacenter access to even more sufficient leading edge GPUs² 3️⃣ sufficiently tight supply of chips (from #1/#2 above) to incentivize domestic design and manufacturing __ ¹x.com/AnthPB/status/… ²x.com/AnthPB/status/… #fabs $nvda

English
1
1
1
430
Joe Weisenthal
Joe Weisenthal@TheStalwart·
If the ingredients to strong AI models are just: Nvidia chips and good engineers, it's not really weird that Chinese open models are near parity. But why hasn't OpenAI's dominant market share translated into reams of unique data, that it can turn into compounding advantage?
English
75
17
519
51.9K
Anthony Bardaro
Anthony Bardaro@AnthPB·
Agreed clarity about competitive vs security threats is warranted, but aren't the overtures not only crystal clear, but also a rule-by-law playbook the Trump administration has (and others have) used repeatedly – this news is a trial balloon intended to have a chilling effect, and if jawboning alone doesn't chill enough, they'll resort to kinetic politicking, which may or may not fail in courts, but will have frozen-over their target in the intervening months/years of legal due process… #kimi #k3 #china #open #lawfare
Ethan Mollick@emollick

We need clarity about what sorts of threats the government is worried about. To what extent is this just intended as an industrial policy & to what extent is it based on a real security risk? The investment going into building on top of Chinese open models is huge, stakes are big

English
0
0
1
116
Ethan Mollick
Ethan Mollick@emollick·
We need clarity about what sorts of threats the government is worried about. To what extent is this just intended as an industrial policy & to what extent is it based on a real security risk? The investment going into building on top of Chinese open models is huge, stakes are big
Andrew Curran@AndrewCurran_

The Trump administration is considering an executive order, and other means, to ban Chinese open-source models within in the United States. Kimi K3 has reignited this debate. Reporting this morning by Axios. Commerce is also considering adding Chinese AI labs to the Entity List.

English
36
17
288
31.1K
Anthony Bardaro retweetledi
Anthony Bardaro
Anthony Bardaro@AnthPB·
/2 …also, I give the "inelastic AI demand" bois a pass because: 1️⃣ INELASTIC IS NOTHING: Part of it is temporal and theoretically nothing is purely inelastic indefinitely, due to substitutes/innovation/budgets eventually inducing price sensitivity (tho empirically some public goods or addictive substances come close) 2️⃣ EXPONENTIAL ADOPTION INDISTINGUISHABLE FROM INELASTICITY: Exponential demand growth in AI can reflect and has reflected both shifting demand curves (new capabilities/use cases) and price elasticity (lower costs per task/token unlock more volume), so the two are not mutually exclusive in early adoption phases – kinda rhymes with my "Spontaneous Regeneration" backfilling in Facebook ads¹ 3️⃣ PUSHING ON A STRING: Osbourne Effect² and other frictions to adoption are non-elasticity factors (neither correlation or causation thereof) – lower prices aren't going to depreciate sunk cost legacy infrastructure³ any faster; teach grandma how to AI; spread the word; etc __ ¹medium.com/adventures-in-… ²x.com/i/status/20301… ³x.com/i/status/15649…
Anthony Bardaro tweet media
Anthony Bardaro@AnthPB

...Jensen's: "our Al biz [has] no real installed base...all brand-new things that people are growing into" ...beckons my: "infrastructure [needs] to get fully depreciated [before] installation and adoption of newer/more efficient/more productive" x.com/AnthPB/status/… $nvda

English
1
1
1
117
Steve Hou
Steve Hou@stevehou·
"The Jevons paradox" is not an economic law. It's an empirical observation just like "the Philips curve". When someone cites "the Jevons paradox" wrt AI, it's much more akin to a prediction that the price demand elasticity of AI is very high or at least higher than 1.
Jared Sleeper@JaredSleeper

I love the tweets that cite Jevon’s paradox as some sort of economic law, rather than an attempt to explain unusual behavior that sometimes occurs but is not the norm. It is a (very useful) framework, but not an argument in and of itself. Many tech cases where it did not hold

English
13
4
80
19.7K
Anthony Bardaro retweetledi
Anthony Bardaro
Anthony Bardaro@AnthPB·
/1 Yes, and nothing wrong with Jevons Paradox being more empirical and historical (inductive) than more theoretical and ceteris paribus (deductive) economic laws – and little wrong with the latter too, fwiw – but the OP minimizing Jevon's because empirical not theoretical is kinda funny in high fallibility social sciences circles (vs hard sciences)… okay, so it's a conditional outcome of (high) price elasticity – economic laws test inductive patterns against deductive frameworks, so nothing wrong with Jevons as synecdoche for both (and seems more like applied economics cum finance cum capital markets/investing tbh)…
English
1
1
1
149
Joe Weisenthal
Joe Weisenthal@TheStalwart·
Something I've been meaning to ask. If you want to a well-attended baseball game these days, how many people in the stadium are still filling out box scores by hand? It's been a long time since I've seen one. But going to Tigers games in the 80s, you'd see am all over the place.
Joe Weisenthal@TheStalwart

Obviously baseball is the ultimate end of one spectrum. It's a game where fans literally used to bring their own spreadsheets (of a type) to the stands, and fill out each event as it went along. Soccer is the other far end. So how do they compute momentum, XG, and so forth?

English
71
1
230
76.3K
Steve Hou
Steve Hou@stevehou·
@AnthPB I was surprised by how big and how late it was
English
1
0
1
967
Steve Hou
Steve Hou@stevehou·
This is a level of precise prescience that I’ve not seen shown by anyone else on X. And I’ve paid relatively close attention. All the “semiconductors specialists with newsletters” really came much later like last year or so when the trade has already happened and caught attention
Steve Hou tweet media
English
9
1
103
17.3K
Anthony Bardaro
Anthony Bardaro@AnthPB·
The funniest outcome is either/both Thinking Machines Labs and Safe Superintelligence (Mira Murati and Ilya Sutskever respectively) publicly releasing standalone AI for general availability – from stealth-ish and pre-revenue to lobbing grenades into the fray… #ai #ssi #neolabs #ami
English
1
0
1
652
Anthony Bardaro
Anthony Bardaro@AnthPB·
cmon, give $okta its flowers – has crawled through mud and come out a phenomenal, quality business (e.g. GMs ~77% and FCF margins 30%+ with a lotta net cash and $1B buyback)… valuation is relatively cheap for a reason, namely growth – like, even adding back +100bps to revenue growth from GSI offboards, it’s still somehow barely a LDD grower, and idk how that’s possible after new products (OIG, Okta/Auth0 for AI Agents, etc) were 25% of fQ1 bookings, with +40% ACV uplift when bundled… but it’s proven itself a quality biz with accelerating RPO (imho), and OIN is a really important repository too fwiw (not nothing)…
English
1
0
2
472
DaRazor
DaRazor@akramsrazor·
$okta ahh yes it became the $twlo for agentic identity bottleneck huh
English
2
1
15
6.7K

Keşfet

@matt_slotnick @TheStalwart @stevehou @elonmusk @BarackObama @taylorswift13 @cristiano @BillGates