Pinned Tweet
Start with something like PopperPad.
PopperPad is scientific memory for humans and agents: an append-only, falsification-first knowledge ledger for hypotheses, recipes, evidence, counterexamples, artifacts, and typed relationships between claims.
github.com/TheDarkLightX/Pop…
When judging character, look at the total person. Look at the good they do for others when they don't have to do it. This is a Richard Heart PSA. Ask your AI to look up other Richard Heart PSAs. How many other founders put out PSAs?
Lets make a more exhaustive list of ways to "get hacked." Many of which I've mentioned before.
Weak RNG:
Use a wallet with a weak RNG (random number generator.) or other vulnerability. Some mobile wallets had this problem, and some vanity address generators had it too.
Someone has a camera watching your screen and you view your seeds.
You google a website and a scam site has done SEO or paid to be at the top of the search results
Fake "support" messages you by direct message, or on socials to help you with your wallet or problem.
Fake support pretends to be the exchange and asks you info they use to login as you and empty you, to "verify your account."
You accidentally leak your seed on a livestream (sounds erotic.)
You put your seed in plain text somewhere, and someone else finds it.
You use a brain wallet with a phrase from a book and people constantly scan the chain for common phrases from books.
You used an L2, and the L2 decided to take your money.
You used an exchange and they decided to take your money.
You used an exchange and they didn't decide to take your money, but got hacked or just exit scammed everyone at once
You installed malware. RAT (remote access trojan). Address replacer, (replaces the address you copied with their address instead of the one you wanted.)
You fell for vanity addresses made to look like one you've sent to in the past, but sent to you more recently, so when you look at the block explorer it looks like a previous legit address, same beginning and end, but the middle is different. People have lost lots of millions to this one recently, heck I think it's the majority of gas use on Ethereum now.
You gave your coins or money to someone else to invest. They lost it / stole it. You fell for a romance scam or pig butchering scam, or AI boss asked me to send money scam or whatever scam of the day is.
You installed an evil browswer extension.
The front end you used got DNS hijacked and now points to an evil dapp.
The X account you follow got hijacked and is now spreading malware links.
You installed an ok browser extension but it got bought by, or exploited by evil and auto updated to evil.
You set too wide a slippage trading on a DEX and got nuked.
The state takes half ur money, cuz, uh, divorce, or whatever reason.
You forget your seed words or don't write them down correctly.
Some guy at the airport security just images your device and decides to empty whatever wallet he finds.
You left a limit order in a wallet with no funds, but then you send funds one day and the stale order fills at a terrible price.
Basically, in computers, physical access defeats most countermeasures, so it's wise to not have any unencrypted seed on any single device in a single place ever.
You approved a dapp's permissions, but then one day the dapp gets evil, often by using an "upgradeable" proxy contract, becuase you never removed the permissions, or overapproved, or jsut shouldn't have ever used a contract wiht an upgradeable proxy ever, anyway.
Oh, yeah, you install malware by doing a job interview, or talking to a reporter, but they're actually just scammers. Devs also fall for this by cloning repo's and installing whatever evil is in them. So the impersonation thing, whether it's for interviews, or investors, or getting hired is a very, very common vector for getting people to install viruses on their machines. People also fall for other kinds of impersonation, people pretending to be their boss, or pretending they need bailed out using AI vids.
You use anything with an admin key.
I could probably think of more, and I've mentioned the majority of these on here before. Feel free to add.
This is an example of Richard Heart's character. He did not have to warn them. He put out the PSA anyway. There are many like these. Make note of the good people do for others.
Ethereum, Solana, Don't open up your metamask for a while, and if it's open, close it. Their infra got hacked. Which means the chances of someone pushing an evil update are probably higher.
I tell everyone how to do things better, but they don't listen, lol. Auto-updates are a huge security risk.
This is why the software at PulseChain, PulseX, HEX, is software you can download and not worry about it autoupdating on you.
Hopefully metamask gets it all sorted, but even if they do, autoupdate will still always be a threat to you.
Just looking out for you all out there.
D.#dwards retweeted
Dear skeptics of multi-provider AI workflows: the world is about to start looking very strange to you.
Specifically, you'll start seeing a lot of math results coming out from independent researchers and groups using models from two or more frontier providers collaborating on the same project, because you can get much better performance that way, compared to using a single provider.
After math, other fields will follow, starting with basic physics and materials science. For each field, first you'll see hype from single-agent and multi-agent teams from a single provider, like a swarm of Claudes or a swarm of GPTs. But the then you'll start seeing better results, at least per dollar, coming from independent teams using models from multiple provders in combination.
These multi-provider breakthroughs won't be advertised as heavily, because no single frontier lab will pay for the marketing. But if you watch for who outside frontier labs is able to crank out impressive results, *without* billion dollar budgets, you'll start to see the pattern.
To many, this multi-provider advantage is obvious in advance, e.g., if you read and grokked my ARCHES report from 2020. Sorry if this post was boring for you!
But if this is *not* yet obvious to you, and especially if you're skeptical, you should probably start updating how you think about intelligence.
D.#dwards retweeted
I keep seeing AI-assisted attacks. It's nothing novel or crazy, just noisy.
A human operator runs a recon command once and moves on. An AI will try anything possible, even if that means running the same recon command 400 times. You also start seeing the patterns, like echo markers wrapped around commands so the agent can parse its own output.
Good news for defenders: the basics still catch it, volume and behavior give it away.
I'm saying it for the 100th time: AI isn't making attackers smarter. It's making them louder and maybe a bit faster.
D.#dwards retweeted
Shaping & Monitoring Models through their Internal Representations
Mechanistic Shaping trains models to restrict targeted internal concepts or behaviors without hurting task performance, while keeping those concepts detectable by refitted probes.
arxivb.org/abs/2611.07071
D.#dwards retweeted
Zyphra's new paper proposes Zyphra Hierarchical Attention (ZHA), which organizes the KV-cache into a tree of compressed summaries. Queries search the tree from top to bottom, narrowing down which tokens deserve attention.
With just 1B tokens of additional distillation training, ZHA preserves most of the performance of full attention on models up to 35B parameters. On RULER at 256K context, it achieves 93.3% vs 96% accuracy, while attending to just 1-6% of the KV-cache.
alphaxiv.org/abs/2610.42112
D.#dwards retweeted
Exactly, without making the training procedures & datasets public, any claims of "emergent" properties in artificial intelligence (AI) solely from marketing claims by AI companies should be taken with a huge grain of salt.
Its not scientific to take the claims as facts.
D.#dwards retweeted
The statistical modeling, used in AI, is pretty much brute-force to some extent because it needs tons of data to model events as sampled from some fixed true distribution P().
That distribution is there in reality but not fixed, it's changing due to complex interactions.
Before trying to solve any problem, define the limits. This shapes the solution space. The Bekenstein bound for example.
So we have huge data centers, lots of space, RAM, storage, to store these increasingly growing models, which I see as fuzzy layered lookup tables. It's very inefficient compared to a brain running on 20 watts, but it's the best we have under classical physics.
Time complexity, space complexity, we have ways to determine if a problem can be solved in reasonable time. It's a tradeoff. You can make a lookup table big but then you lose space. You can keep it small, now you lose time. Modern AI hasn't changed this.
The gamble is time investment. You don't have infinite time. Some problems take more time than our universe has. Some problems take more atoms than our universe has. Picking problems well, is the main skill of mathematicians, programmers, and scientists.
Einstein for example spent years on E=MC2. There were hints it was tractable, from the logic side, but many people invest their whole life into a problem which they never solve. He solved it in finite time.
The same applies to people, each person has a finite amount of time to solve any problems in their life. And each problem takes some time. Hard are problems which cost a lot in time and resources for a human to solve.
Easy and hard, are really just due to not having enough resources to find a solution. Any solution which you have to solve brute force, and that takes intractable amounts of time, is hard. AES 265 is hard, it's intractable to brute force it.
If we had an efficient search algorithm, we'd have ASI. Give a problem, and with enough data and compute, it will find the grain of sand in the seemingly endless desert of candidate solutions. LLMs can't do this. And logical AI can't do this. Both aren't efficient enough.