David Sacks retweeted
So the Ukraine war could have been over in 2023 but Biden and the West rejected it
I confirm the 2023 peace-treaty plan that was leaked to the FT.
What the draft envisaged:
- Russia withdraws its troops from most of the four oblasts it had just declared “annexed.”
- The new line of demarcation runs roughly along the line as of February 24, 2022: Crimea and the part of Donbas that Russia controlled before the full-scale invasion.
- Ukraine recognizes Russian control over Crimea.
- Ukraine grants Russia an unobstructed land corridor to Crimea.
- The ceasefire regime is guaranteed by NATO countries that provide protection for Ukraine.
- Russia begins negotiations with the US on its own security guarantees.
From my side, I’ll add a few insider details that few people know about.
The plan was discussed NOT ONLY by Rodnyansky: the key figure was ABRAMOVICH.
Moreover, Russia was ready to give up the Zaporizhzhia Nuclear Power Plant as well. And the 300 billion in frozen assets was to go toward rebuilding all of Ukraine, that is, together with the occupied territories.
As for Crimea: we were not required to recognize it as Russian, only to recognize Russia’s control over it. In other words, we acknowledge that right now it is under Russian control (which is how it is). We were also required to adopt a law banning the return of Crimea by military means.
The land corridor to Crimea, a highway and a railway, was to be built at Russia’s expense. The hardest question was how to guard it. That is why the Russians were ready to give up Melitopol, but not Tokmak: they feared we would cut the corridor with artillery.
Zelensky rejected this proposal at the time.
186
790
5,468
148,949
David Sacks retweeted
POD UP! 🚨
-- AI Cyber Panic w/ @nikesharora
-- Ken Griffin’s Bubble Warning
-- The Fall of France?
(0:00) Bestie intros!
(0:11) Ken Griffin's AI bubble warning
(8:19) Palo Alto Networks CEO Nikesh Arora joins
(14:24) AI Cyber panic: PHP exploits and the tumultuous year n the brink: Riots, immigration referendum, debt crisis
(1:31:49) Trump's second half strategy, Press Secretary J-Cal?
-- AI Cyber Panic w/ @nikesharora
-- Ken Griffin’s Bubble Warning
-- The Fall of France?
(0:00) Bestie intros!
(0:11) Ken Griffin's AI bubble warning
(8:19) Palo Alto Networks CEO Nikesh Arora joins
(14:24) AI Cyber panic: PHP exploits and the tumultuous year n the brink: Riots, immigration referendum, debt crisis
(1:31:49) Trump's second half strategy, Press Secretary J-Cal?
70
68
573
176,786
David Sacks retweeted
NEW: Ken Griffin on Paris riots, Chicago and JB Pritzker
I think Paris should be a real wake up call to every single young American. France is the dream the left talks about. That dream is the nightmare I left from Chicago. That dream doesn’t belong in our country. A society much you can take from others.
Do you see candidates that you’re actually fearful that’s the path they’re going down:
J.B. Pritzker, Governor of Illinois. Yes, 100%. 100%
I think Paris should be a real wake up call to every single young American. France is the dream the left talks about. That dream is the nightmare I left from Chicago. That dream doesn’t belong in our country. A society much you can take from others.
Do you see candidates that you’re actually fearful that’s the path they’re going down:
J.B. Pritzker, Governor of Illinois. Yes, 100%. 100%
Ken Griffin on the rise of DSA, Paris riots and looting in Chicago
I'm not so worried about it. It's a New York, Illinois, Michigan, California bit of a bubble.
If you want to see where socialism leads to, look at France last weekend.
It's the socialist dream. The country e government level.
BTW Paris looked pretty damn similar to Chicago in 2021-22.
It’s a people… (that) are very willing to riot and protest anything. Is that really the future that the US might be looking at?
It’s the future. I lived in Chicago.
I could look out my window at the Dior store. They were very efficient.
We may be more efficient than the French, even at looting.
95
921
4,556
287,959
David Sacks retweeted
What’s actually going on in Ukraine is that there is a massive incentive difference between the elites and the people.
The former have grown very rich off the hundreds of billions flowing to Ukraine, while the latter’s children are conscripted to die in trenches.
The former have grown very rich off the hundreds of billions flowing to Ukraine, while the latter’s children are conscripted to die in trenches.
25,874
42,814
255,861
13,820,918
There’s a world of difference between user-authorized agents (Muse, Instinct, Grokbot, etc.) and true unauthorized/rogue agents that could constitute cyberattacks.
Most of the activity described in the Axios story appears to be the former: customers sharing credentials with agents so they can take actions on their behalf. This may violate bank terms requiring humans to log in themselves, but these are in no way cyberattacks.
The direction banks should go is to support user-directed agents. For example, Amex is an authorized connection in Muse, making it easy for users to grant, limit, and revoke access.
Read-only permission levels are a sensible interim step. Several banks already support or are moving toward them. Full write access (with user credentials) is a different risk category and should be approached carefully.
Frightening the public by blurring the line between user-authorized agentic behavior and cyberattacks is irresponsible.
Most of the activity described in the Axios story appears to be the former: customers sharing credentials with agents so they can take actions on their behalf. This may violate bank terms requiring humans to log in themselves, but these are in no way cyberattacks.
The direction banks should go is to support user-directed agents. For example, Amex is an authorized connection in Muse, making it easy for users to grant, limit, and revoke access.
Read-only permission levels are a sensible interim step. Several banks already support or are moving toward them. Full write access (with user credentials) is a different risk category and should be approached carefully.
Frightening the public by blurring the line between user-authorized agentic behavior and cyberattacks is irresponsible.
SCOOP: Major U.S. banks are detecting unauthorized third-party AI agents inside customer accounts, including some attempting to take actions that appear to be independent of user instructions, sources told Axios.
The findings have alarmed banking executives, who are racing to restrict access to their systems as AI moves at a velocity faster than financial institutions can respond, multiple industry and bank executives told Axios.
Several banks are now preparing for worst-case and even “day after” scenarios, including how to rapidly secure or restore customer accounts if things go wrong, the people said.
Several bank executives told Axios they're struggling to distinguish between human users and AI agents. One executive said banks also struggle to determine whether agents are authorized by customers or potentially operating without their permission, more akin to rogue agent behavior.
Read my latest for Axios here: axios.com/2026/10/22/ai-…
378
726
5,823
362,942
You’ve put 20% of your equity in an LLC. Whether it serves the public good remains to be seen.
Between this and our prior commitments, roughly 20% of all Anthropic equity will be dedicated to the public good, and that commitment can grow as Anthropic grows.
We plan to use these funds to ensure more people can share in the benefits of AI, and to increase our ability to provide funding to deal with its risks.
267
289
5,994
291,272
David Sacks retweeted
@AnthropicAI I recommend giving the shares directly to the public
1,390
1,837
42,107
2,160,138
David Sacks retweeted
Sen. Elissa Slotkin (D., Mich.) just gave an interview where she said the quiet part out loud. She intends to unleash a slew of investigations to deter young people from even considering supporting the Administration … if they know what is good for them. jonathanturley.org/2026/10/18/i-w…
1,760
7,936
22,633
451,235
David Sacks retweeted
David Sacks on All In: payback period as the Northstar metric for AI infrastructure.
CoreWeave says typical five year deals deliver about 2.5 years payback. SpaceX pays back in roughly one year because it monetizes around $50B per gigawatt while CoreWeave sits near $20B per ffectively spot pricing on 90 day cancelable deals, while CoreWeave locked five year contracts at older prices and spot compute moved way up on scarcity and demand. As long as payback stays short, the economics are nearly unlosable and debt rolls fast. The warning signal is payback stretching out, which implies demand softening, spot falling, or execution delays.
CoreWeave says typical five year deals deliver about 2.5 years payback. SpaceX pays back in roughly one year because it monetizes around $50B per gigawatt while CoreWeave sits near $20B per ffectively spot pricing on 90 day cancelable deals, while CoreWeave locked five year contracts at older prices and spot compute moved way up on scarcity and demand. As long as payback stays short, the economics are nearly unlosable and debt rolls fast. The warning signal is payback stretching out, which implies demand softening, spot falling, or execution delays.
17
37
363
57,454
This illustrates the point I’ve been making about the limits of centralized approaches to SI safety in cyber.
OpenCode (and earlier Hugging Face) hit the same wall mid-attack: frontier models refused to analyze the very exploits needed for defense because the payloads tripped safety filters. The guardrails backfired.
Meanwhile, the attackers found a way around restrictions to access advanced capabilities. This is not surprising. Chinese models are only about 6–9 months behind the American frontier in cyber. Obviously the U.S. has no jurisdiction over them.
The only durable response is decentralized: make powerful models available so defenders can build and run their own cyber defenses. The only solution to SI-powered cyberattacks is SI-powered cyber defense.
OpenCode (and earlier Hugging Face) hit the same wall mid-attack: frontier models refused to analyze the very exploits needed for defense because the payloads tripped safety filters. The guardrails backfired.
Meanwhile, the attackers found a way around restrictions to access advanced capabilities. This is not surprising. Chinese models are only about 6–9 months behind the American frontier in cyber. Obviously the U.S. has no jurisdiction over them.
The only durable response is decentralized: make powerful models available so defenders can build and run their own cyber defenses. The only solution to SI-powered cyberattacks is SI-powered cyber defense.
OpenCode is currently under a sustained cyberattack, and mid-incident they ran into the same problem Hugging Face did: the frontier models they were using for defense began refusing to help fight attacking agents because analyzing the attacks started triggering safety guardrails.
201
356
2,818
251,336
Is 10²⁵ FLOPs Still the Frontier?
A new Anthropic-written safety bill is circulating on Capitol Hill. Some of its ideas could be reasonable if limited to true frontier companies. As written, however, it will eventually sweep in most of Silicon Valley, damaging the startup ecosystem.
The bill defines the “frontier” as any model trained on 10²⁵ FLOPs — the EU standard from 2023. Biden’s executive order, California’s SB 53, and New York’s RAISE Act used 10²⁶ (ten times more). 10²⁷ is where the frontier actually is today (e.g., GPT-6 Astra).
Astra was reportedly trained on 100,000+ Grace Blackwell GPUs. A 10²⁵ model can be trained on as few as several hundred GB300 chips in three months. That is not frontier in 2026.
The definition is also cumulative: it adds FLOPs from fine-tuning, reinforcement learning, or other material modifications. In practice, that captures almost all meaningful post-training — work that is becoming commonplace in the enterprise.
The enterprise trend is toward thousands of smaller custom models built for specific workloads. These pose no frontier risk. Treating them as frontier systems would impose large compliance costs across corporate America for no safety gain.
Dario has claimed that Anthropic designs rules to disadvantage frontier companies like his own while advantaging smaller competitors. This definition does the opposite.
A stale 2023 cutoff favors incumbents who already clear the bar and can absorb the costs. It hobbles the open model ecosystem that threatens Anthropic’s pricing power.
The frontier keeps moving. Any definition meant to track risk should move with it.
Anthropic should clarify if they really believe 10²⁵ FLOPs represents the frontier.
A new Anthropic-written safety bill is circulating on Capitol Hill. Some of its ideas could be reasonable if limited to true frontier companies. As written, however, it will eventually sweep in most of Silicon Valley, damaging the startup ecosystem.
The bill defines the “frontier” as any model trained on 10²⁵ FLOPs — the EU standard from 2023. Biden’s executive order, California’s SB 53, and New York’s RAISE Act used 10²⁶ (ten times more). 10²⁷ is where the frontier actually is today (e.g., GPT-6 Astra).
Astra was reportedly trained on 100,000+ Grace Blackwell GPUs. A 10²⁵ model can be trained on as few as several hundred GB300 chips in three months. That is not frontier in 2026.
The definition is also cumulative: it adds FLOPs from fine-tuning, reinforcement learning, or other material modifications. In practice, that captures almost all meaningful post-training — work that is becoming commonplace in the enterprise.
The enterprise trend is toward thousands of smaller custom models built for specific workloads. These pose no frontier risk. Treating them as frontier systems would impose large compliance costs across corporate America for no safety gain.
Dario has claimed that Anthropic designs rules to disadvantage frontier companies like his own while advantaging smaller competitors. This definition does the opposite.
A stale 2023 cutoff favors incumbents who already clear the bar and can absorb the costs. It hobbles the open model ecosystem that threatens Anthropic’s pricing power.
The frontier keeps moving. Any definition meant to track risk should move with it.
Anthropic should clarify if they really believe 10²⁵ FLOPs represents the frontier.
162
218
1,924
170,502
👇
@DavidSacks $META's Llama 3.1 405B was pre-trained on 3.8×10^25 FLOPs and has been publicly downloadable since July 2024. A 10^25 cutoff classifies a two-year-old open-weight model as frontier.
4
7
123
31,495
👇
@DavidSacksive definition is the part that would actually hit me. Post-training — fine-tuning, RL — is the whole job now. I do that on small models every week, and if the FLOPs all stack against the 10^25 line, I'd be filing compliance paperwork for models that are nowhere near frontier.
5
1
82
22,948
David Sacks retweeted
David Sacks pushes back against the prevailing narratives, arguing that CoreWeave’s sharp spike in bond yields and stock selloff reflect market volatility driven by bets on multi-trillion-dollar AI outcomes and exponential growth, rather than a fundamental collapse in demand. He breaks down these narratives one by one.
The key to understanding CoreWeave’s credit risk is that the creditworthiness of its long-term contract customers matters far more. x.com/genzzztrade/st…
The key to understanding CoreWeave’s credit risk is that the creditworthiness of its long-term contract customers matters far more. x.com/genzzztrade/st…
16
27
284
50,695
David Sacks retweeted
🚨 David Sacks Says the Timing of Anthropic’s Small Model ‘Scare Campaign’ is ‘Very Suspicious’
“This is where the enterprise is headed right now … You're starting to see this trend where small models are taking share, and lo and behold, coincidentally Anthropic is now starting t small models.”
“This is where the enterprise is headed right now … You're starting to see this trend where small models are taking share, and lo and behold, coincidentally Anthropic is now starting t small models.”
19
60
417
43,622
David Sacks retweeted
POD UP! 🚨
Ben Horowitz (@bhorowitz) fills in for @friedberg!
-- CoreWeave's Tanks After Debt Selloff: Did the AI Trade Just Crack? $CRWV
-- AI Deflation and the Future of Software
-- Anthropic IPO: Red Flags, Valuation, How Much to Raise?
-- Psyop Corner with Sacks
-- What’s Behind His AI Warnings?
(0:00) Ben Horowitz fills in for Friedberg!
(2:18) CoreWeave's debt selloff: did the AI bubble just pop?
(31:14) AI's deflationary impact: decompiled software
(45:44) Anthropic's investor day gets mixed reactions
(1:03:45) Psyop Corner: Anthropic warns about self-replicating models on IoT devices
(1:22:19) Understanding Bill Gates the Doomer
Ben Horowitz (@bhorowitz) fills in for @friedberg!
-- CoreWeave's Tanks After Debt Selloff: Did the AI Trade Just Crack? $CRWV
-- AI Deflation and the Future of Software
-- Anthropic IPO: Red Flags, Valuation, How Much to Raise?
-- Psyop Corner with Sacks
-- What’s Behind His AI Warnings?
(0:00) Ben Horowitz fills in for Friedberg!
(2:18) CoreWeave's debt selloff: did the AI bubble just pop?
(31:14) AI's deflationary impact: decompiled software
(45:44) Anthropic's investor day gets mixed reactions
(1:03:45) Psyop Corner: Anthropic warns about self-replicating models on IoT devices
(1:22:19) Understanding Bill Gates the Doomer
71
79
885
337,380
Patrick is one of the sharpest people I’ve worked with in government. No one knows crypto policy better or worked harder to get market structure done. He negotiated the compromises needed to get Clarity out of committee and onto the Senate floor — without him it wouldn’t have gotten that far. The country owes him a debt of gratitude for his service at the White House, the Pentagon, and in uniform. Looking forward to seeing what he does next in the private sector. 🇺🇸
Serving in this administration has been the privilege of a lifetime.
From helping to stand up the most successful presidential transition in history alongside @howardlutnick and @EDSecMcMahon, to rebuilding our defense industrial base as PTDO Deputy Under Secretary for R&E and then Acting Director of OSC at @DeptofWar, to driving forward President Trump’s vision to make the United States the “crypto capital of the world” as Executive Director of the @WhiteHouse Crypto Council, I consider myself to have been the luckiest guy in the world over these past two years.
First and foremost, I want to thank @POTUS for the opportunity to serve the American people again in your second term. Your clear-eyed vision to ensure the United States remains the undisputed global superpower at this time of unprecedented technological change and geopolitical strife is critical to our nation’s future prosperity.
Second, I want to thank @DavidSacks for your visionary leadership and mentorship. I am so grateful to have been able to work with you and learn from you over the past year. Thank you also to @USWREMichael, James Mazol, @HarryYJung, Nick Elliot, Tyler Williams, @lukejpettit, Jonathan Hurowitz, Andrew Rittenhouse, Michael Friedman, @atitus7, @passalacqua_mj, @willscharf, @JamesCBraid, @MayMailman, Steve Levine, Patrick Butler, Zach Carstens, and countless others across the administration for being the best colleagues anyone could ask for. It was an honor to serve alongside you.
Lastly, thank you to @SECPaulSAtkins and @ChairmanSelig, @SecScottBessent, @USComptroller Gould, @SenatorTimScott, @SenLummis, @SenatorHagerty, @berniemoreno, @JohnBoozman, @SenThomTillis, @RepFrenchHill, and @RepBryanSteil for your partnership in advancing crypto policy over the past year with the passage of the GENIUS Act and our work to advance the CLARITY Act. With regulators and elected officials like you, I rest easy knowing the future of digital asset innovation in America is bright.
Regarding next steps, I’ll have to more to share soon. For now, I’ll say this: I am extremely bullish on America as we embark upon a new golden age of technological advancement across multiple frontiers. I’m grateful to have done my part to unleash American innovation while in government, and I look forward to advancing that same mission in the private sector! 🇺🇸
66
283
4,065
199,921
David Sacks retweeted
@DavidSacks The whole thing was fake. Not only was the “hacking” trivial it was done by a GLM but then when it’s “self-replicating” they swap out the model to a small Llama because the GLM is too big 😂
9
14
208
62,584
David Sacks retweeted
@FedorovMykhailoms are owned and controlled by the US government. Any limits on speed or usage are decided by the US military and ultimately POTUS.
Starlink is a commercial service owned and controlled by SpaceX. Usage of Starlink is limited to civilian or defensive activities and cannot be used for offensive military action, as doing so would be inappropriate for a commercial company.
Starlink is a commercial service owned and controlled by SpaceX. Usage of Starlink is limited to civilian or defensive activities and cannot be used for offensive military action, as doing so would be inappropriate for a commercial company.
3,744
6,916
53,450
4,666,076