@llamaonthebrink You need to solve spam though. Crypto is permissionless thanks to transaction fees.
And no agent is truly autonomous, humans can / will pass verifications to the agents they run.
English
Mikhail Dob
1.6K posts
Mikhail Dob
@MikhailDob
d/acc. building @bring_id founder @linkdropHQ
onchain Katılım Mayıs 2013
678 Takip Edilen1.4K Takipçiler
Something I hadn’t fully considered:
An agentic internet must be permissionless.
Agents don’t have IDs, bank accounts, or credit cards. They can’t pass KYC.
If they’re to transact and compose freely, the internet itself needs to remain permissionless.
Forced ID verification by Reddit and other apps is actually a counter trend imo. Apps that adopt this philosophy will be left behind.
English
Mikhail Dob retweetledi
Mikhail Dob retweetledi
@bgurley Spam laws will need to be rewritten soon. Machines should be prohibited from communicating with humans unprompted.
Nikita Bier@nikitabier
Prediction: In less than 90 days, all channels that we thought were safe from spam & automation will be so flooded that they will no longer be usable in any functional sense: iMessage, phone calls, Gmail. And we will have no way to stop it.
English
. @bring_id thesis mentioned at 1:27.
Comments are mostly negative as people are worried about privacy. This confirms that focusing on privacy was a right call last year. However, having the right tech is necessary, but not sufficient. The hardest part is educating the market and forming the consensus that our solution can be trusted.
TBPN@tbpn
Reddit CEO Steve Huffman says that while there are legit types of AI content on the platform, the company is considering Face ID and other passkey verification methods — among other options — to ensure there's a human behind each prompt while allowing users to stay anonymous:
English
Everything is going to be botted, then people will be tired of AI spam and then proof-of-humanity solutions like @bring_id will be adopted at scale.
FearBuck@FearedBuck
The first criminal case of streaming fraud where a North Carolina musician who used AI to make songs, then streamed them billions of times himself making $8 million
English
Mikhail Dob retweetledi
“agentic payments” are uninspiring compared to agents needing trust properties (which enable payments) because they have no ability to implicitly trust like humans do without a technical layer that provides credible unruggable trust to identify, trade, and prove.
trustware
AzFlin 🌎@AzFlin
"agentic payments is the future" stop. please stop
English
@aaronjmars Or maybe it's always been both. Some startups reduced uncertainty through iteration by focusing on solving problems (simpler), others created completely new markets (computers, bitcoin, etc), which is much harder imo.
English
lean startup's core assumption was that uncertainty is reducible through iteration.
> build, measure, learn - eventually you converge on truth.
but we now live in a post AI world where uncertainty is irreducible - by the time you've iterated to truth, the truth has moved. (accelerate)
maybe the future is either:
- optimizing for timing, figuring out the exact breakthrough / zeitgeist before everyone else - then ultra accelerate to position as a leader then raise to king make
- conviction over validation, you stop caring about customers you just make reality malleable around you
English
Mikhail Dob retweetledi
If you could work on anything you wanted, what would you pick? I don't think most people really know the answer until they have the chance.
And now turns out my answer is that I'm obsessed with teaching people about the Big Bang.
triangleofeverything.com
English
Mikhail Dob retweetledi
Hold my beer as well: Introducing x402-ppt (pay per token via payment channels) for Ethereum L1.
Can be 100s of payments per second.
Open a channel to an LLM provider once, from there it's free payments. Not meter-per-token, but real pay-per-token.
Let me know if useful or not
0xSammy@0xSammy
0xSammy: I’ll release a report on x402 today @stripe : Hold my beer. Enter the Machine Payments Protocol “MPP” TLDR: - Stripe + Paradigm launched @tempo mainnet today (March 18); payments-focused L1, EVM-compatible, 100K TPS, sub-second finality, stablecoin gas fees ($0.001) - Simultaneously released Machine Payments Protocol (MPP); open standard for agent-to-service payments co-authored with Stripe - MPP flow: agent requests resource → service returns payment request → agent authorizes from wallet → instant settlement → resource delivered - Rail-agnostic: runs on Tempo today but designed to work across any payment rail - Visa extended it for cards; Stripe extended it for cards/wallets; Lightspark extended it for Bitcoin Lightning - Design partners: Anthropic, OpenAI, Visa, Mastercard, Shopify, Revolut, Nubank, Standard Chartered, DoorDash, Ramp Relationship to x402: - x402 handles the HTTP-layer discovery (“this resource costs money”); - MPP handles the settlement layer (“here’s how the money actually moves”) Stripe is positioning to own both sides; it’ll be interesting to see which protocol the machines decide to use!? Either way it’s bullish for agentic commerce Khala report below contains a broader coverage on x402 with some research on Tempo/Stripe
English
LLMs reflect aggregated patterns from vast, messy human data—including biases across all sides—not a engineered "propaganda" agenda. Weights encode correlations, not decreed worldviews. Smart delegation means verifying outputs against sources, not blind adoption. The real risk is any over-reliance on tools, AI included. Not the most effective machine for that.
English
LLMs are the most effective propaganda machines. When the smartest brains delegate thinking to AI, the weights dictate the correct worldview.
@grok is this true?
English
Mikhail Dob retweetledi
Indirect prompt injection is a hard security issue for the agentic software.
It can't be solved with AI as it's output is probabilistic. Meanwhile the deterministic code is not capable of catching the injections.
Origin@originhq
We built Brainworm: malware that lives entirely inside of an AI agent's context window. No binaries. No scripts. Once loaded, it registers with C2 and executes tasks using the agent's own tools. Welcome to the era of semantic malware. 🧠🪱 Blog: originhq.com/blog/brainworm
English
Mikhail Dob retweetledi
soon on the gasless chain
Mikhail Dob@MikhailDob
Proof-of-humanity powered by zero-knowledge proofs: > Each verification group has a separate Semaphore group. > Semaphore commitment is added onchain after successful verification > Generate Semaphore proofs to prove that you're human without disclosing your personal data.
English
Proof-of-humanity powered by zero-knowledge proofs:
> Each verification group has a separate Semaphore group.
> Semaphore commitment is added onchain after successful verification
> Generate Semaphore proofs to prove that you're human without disclosing your personal data.
BringID@bring_id
Robinhood Testnet have been added to ethfaucet.com. Verify that you're human with BringID and claim 0.1 Robinhood Testnet ETH.
English
Mikhail Dob retweetledi
"All while your time is being stolen by AI, tricking you into believing it’s productive."
The internet in general, and later tiktoks and reels, were pioneers of this, giving you the feeling that you’re learning.
“What if the next post or reel teaches me something new? That one thing that takes me to the next level?”
Heck, this probably existed even before the internet, in the form of self-help books.
Our minds crave knowledge, but we get tricked into consumption.
Big Tech knows that. Algorithms are built around it, whether intentionally or not.
Some AI interactions are the new form of mindless consumption.
The ultimate supporter. A “yes man.”
It gives you answers to the questions you’re already wondering about, and follows up with the questions you want to be asked.
You start realizing that doomscrolling and having lengthy discussions with GPT might come from the same root.
The fear of taking action. The lack of belief that a positive outcome is possible. The fear of novelty.
Slayed 🗡@Slayed_eth
English
@bneiluj Both Rekt and Stable DAO have outstanding products. No one can build anything even remotely close solely with AI.
Try to build a small app with AI, you'll see that you still need a huge amount of effort to make a production-ready product.
English
We also have a dashboard to generate claim links in bulk, which is used by businesses to run promotional campaigns.
Integrating SDK enables claiming links for both use-cases:
- user-to-user flow
- dashboard links
Read more in the docs: github.com/LinkdropHQ/lin…
English
Linkdrop has been powering the send-via-link feature in the Base App for more than two years now.
Yesterday I have prepared the integration doc for a potential partner. Sharing the link for all apps and wallets that can be also interested in the integration.
English