Father of Danksharding | Researcher @tempo | Prev. Founder @caracarehealth & Researcher @ethereumfndn | Let's build accessible blockchain finance for all!
Earth
- Tweets 5,988
- Following 527
- Followers 35,554
- Likes 3,709
Dankrad Feist retweeted
INTRODUCING MERCATOR by @tempo.
Mercator gives agents a new effort level.
By giving an agent a search engine to the world's paid tools it empowers a small model to punch above its weight, while being cheaper than larger models.
$5 Mercator credits w/ your X/Github ⬇️
26
32
274
60,566
Update your iPhones today!
@Apple Finally put passkeys into the secure enclave ONLY! They should not be decrypted outside the enclave. They can, and will, be stolen...
Zero-day remote code execution vulnerability in iPhone Safari. Click a link, and your crypto, passwords and everything else on your iPhone are gone.
Exploited in the wild by "DarkSword" malware.
"The DarkSword attack program has leaked, with its core capability being: extracting forensic-level data from iOS devices via HTTP interfaces. In actual attacks, attackers can combine social engineering or watering hole attacks to lure users into falling victim, thereby stealing data from iPhone / iPad devices and uploading it to servers controlled by the attackers."
Update iPhones immediately. Apple originally patched this, but rumours suggest even the latest versions are vulnerable, “pending confirmation,” across a wider range than the original 18.4–18.7 window.
From a Chinese security researcher, SlowMist CISO, @im23pds
nitter.net/im23pds/status/2101265…
8
36
226
48,663
Most people who claim that models aren't dangerous, or that the only way forward is max acceleration without brakes, probably believe this because they are even more scared of the alternative: If we admit that AI is dangerous then it needs to be regulated/controlled, and that means some insiders will have more powerful models (even more than is the case now) than everyone else.
This dilemma isn't really new: Regulation almost always benefits the incumbents, which is bad for everyone else to some extent, but we do it because the alternative of regular food poisoning and planes crashing out of the sky is worse.
The reason so many people look for an ulterior motive for the AI labs asking to be regulated is that they don't grasp that models could be dangerous. But if you try assuming models are getting dangerous, or at least unpredictable, everything falls into place.
3
3
44
5,845
Yudkowskyanism is clearly bullshit and based on black-and-white binary thinking: AI is like god, goes foom, wants to kill all humans
It is also morally repulsive: doesn't matter if billions of people die because extinction risk is the only thing that actually matters, pivotal acts
BUT
I still think the dangers from AI are massive and to be taken extremely seriously. Even if AI doesn't have any intent to kill humans, it can put us all out of jobs, marginalize us and take our resources away (maybe that wheat field is better for solar cells) and lead to our eventual demise. Or it can be used by malicious humans to inflict misery on many.
I do think that there is a real process to get to a good outcome as long as many of the people creating and training it have good intentions. But the chance of it going wrong is quite significant, 10% seems at least a reasonable estimate.
7
1
26
3,367
Dankrad Feist retweeted
A few thoughts on ZCash (disclosure: Paradigm is an investor in ZEC and ZODL):
- All blockchain ecosystems struggle with long-term funding, especially of public goods, and an inflation-funded developer fund is an elegant mechanism to do so. How this fund is allocated is a fair debate, and we should strive to improve it. But, we believe the dev fund is important, particularly in this age of AI cyber capabilities, rate of quantum progress, etc.
- As ZCash grows in acceptance and adoption as a private complement to Bitcoin, we think pure coinvoting governance introduces unpredictability that could limit long-term trust as a monetary asset, and it is better for the project to combine coinvoting with other forms of governance.
107
101
1,047
403,191
Dankrad Feist retweeted
Hiring for product engineers to work on the front lines with us @tempo and our customers
- build protocols (EVM for payments)
- vertically integrated tech stack
- real world products with 100M+ users
- serious customers (fintechs, banks)
- unlimited AI budget to move fast
15
13
214
37,964
Why the ends don't justify the means 101
remember, if you believe that only you can save humanity from extinction, and you are a consequentialist, you are morally willing to commit unspeakable evil in support of that goal.
if your p_doom is 10% in a decade that means your current expected value of the status quo is negative 800 million lives.
so it's completely rational – if you really believe this – to sacrifice single or double digit millions in pursuit of that goal.
this means: if you think the frontier is insufficiently focused on safety, that it's acceptable to release a false flag bioweapon, agent swarm on grid infrastructure, etc. what are a couple million lives in exchange for billions?
this is why EAs and hardcore consequentialists that are "safety pilled" are extremely dangerous.
19
3,703
If we prove P=NP, it's most likely that it will have zero implications for cryptography and instead mean that our mathematical axioms are bad.
Similar thing to Banach-Tarski. Very interesting to the mathematician, but zero implications in the real world. What it really shows us is that nonconstructive mathematics doesn't provide the best models for the physical world.
“My main point, however, is that I don't believe that the equality P=NP will turn out to be helpful even if it is proved, because such a proof will almost surely be nonconstructive.”
— Donald Knuth
30
31
375
62,489
Great. Now we should declare distillation fair use.
SITUATION DETECTED: The US government has sided with OpenAI against the New York Times. The DoJ says training an LLM on copyrighted works does not violate copyright law, and that treating it as infringement would hurt US science, prosperity, and national security.
3
2
39
4,411
Ethena is making progress on the token/equity misalignment problem! I hope we will see many more projects taking similar steps.
Token/equity misalignment has been a killer for crypto investment: Tokens traditionally have zero legal rights, compared to equity which comes with a strong legal claims. When in doubt, equity wins, and token investors end up empty handed. This is a step in the right direction.
We are excited to announce four updates regarding the Ethena ecosystem, further details on each point are provided in the blog linked below:
1. Buyout of early investors:
The Ethena Foundation executed a buyout of all locked tokens from certain major seed investors that sold any ENA within the last 9 months.
2. Alignment of Token & Equity:
The Ethena Foundation and Ethena Labs have reached agreement on a Master Framework Agreement, whereby IP and ownership of value accrued by the protocol is assigned to the Foundation exclusively and governed by token holders with no residual cash flow due to equity investors in the Labs entity.
3. Revenue Buybacks:
Governance proposal now live for the implementation of the fee switch whereby net revenue accrued across all business lines under the Ethena brand will be used to programmatically buy back the ENA token.
The vote for revenue buyback fee switch implementation is now here, and has already been approved by the Risk Committee:
snapshot.box/#/s:ethenagover…
4. Removal of monthly VC unlocks:
The Ethena Foundation and lead investors have agreed to eliminate future overhang associated with monthly VC investor unlocks by releasing unvested tokens. All team tokens remain locked per the original vesting schedules.
Further details and documentation is provided in the blog linked below:
5
9
74
10,630
Any software that still has the same bugs it had a couple of months ago is now a massive red flag: Since the maintainers clearly aren't using AI to rapidly fix everything as they should, they are probably also not fixing all the security holes that AI can now discover in time.
6
4
60
5,002
Dankrad Feist retweeted
IMO both sides of that chat have entrenched long ago, are optimising for different priorities, and ultimately talking past each other
- "decrease" wants to fix the curve to protect ETH's fundamental monetary characteristics (this is my view). we chose the curve years ago without the rigor necessary for such a decision, hindsight tells us a lot
- "keep it the same" wants to protect solo stakers and the heterogeneity of the network
i cant help but point out, with the current curve, solo stakers are down to .3%. from 17% at beacon genesis - they will be a vanishingly small % of the network in the near future. this is a structural problem that keeping the current curve will not fix. if solo stakers and issuance actually had a direct relationship, we should be arguing to increase issuance. it's obviously more nuanced than that
that being said, maintaining the ability for consensus activities w/ consumer hardware/residential bandwidth is still an admirable benchmark to aim for where we can (EIP 7870).
~~~~~
whatever your position: with respect to the people organizing, engaging in good faith discussion: contentious decision-making through endlessly-circular tg chat is NGMI. you have no idea how many people in the chat are actual stakeholders, to what degree, and anyone can DOS the conversation out of all proportion to their stake in the system's future. the venue is completely ill-suited to the question - we need robust, legitimate venues to build consensus among core devs, eth holders, and users of the network
for ~2 years there's been a telegram group where ethereum's monetary policy gets debated.
justin drake started it march 2025, the day after solana's issuance cut failed a vote.
2,715 messages later, three camps that formed in the first 48 hours haven't moved:
1. reduce issuance (anders' 0.5% max-issuance curve, because the network overpays for security and staking slowly eating raw eth is a real risk).
2. don't you dare (hanniabu's math: cut it and coinbase survives on loss-leader economics while the solo staker dies).
3. it's not urgent (sassal: the merge already cut issuance 90% and eth/btc still bled).
meanwhile staking has crossed 1/3 of all eth and keeps grinding up ~50k/day.
but the data shows it flowing to coinbase, binance, kraken, and treasury companies.
everyone agrees the current curve was suboptimal. nobody agrees it's urgent, safe to change, or fixable without breaking something they care about more.
this is what monetary policy looks like with no central bank. the hard part is getting a decentralized system to agree that changing the rules is safe.
12
6
85
17,270
Some time in the future, we will understand that risk free yield on a digital asset is a bug, not a feature (yes this includes USD)
25
8
122
13,826
This is a good time to remind everyone that mathematics is a language that happens to be good at describing nature, among other things.
Mathematical theorems are equivalences: They say that certain ways of saying things in the language of mathematics look different but they are actually the same.
It looks like we are soon reaching the point where LLMs are better at proving them than the best humans. I think that's a very significant step (but likely not the final one towards human equivalent intelligence in all fields)
10
7
101
10,028
Pacing the frontier is the best thing we can do right now IMO. It's a hard coordination problem but still solvable, maybe about 20-30 entities need to commit.
Models are now definitely at the threshold of becoming dangerous. Allowing the defense to catch up will lead to a less disruptive transition.
5
22
4,532