krisoshea.eth

I like blocks | Dev @OlenaProtocol | @ethereum Protocol Fellow | Prev: @SumeroApp

[krisoshea.eth @krisos]

This has been our main obsession @OlenaProtocol for ~2 years, so it's surreal to finally say that registrations are open.

The core problem we've been trying to solve is how to get accurate crowd predictions on questions that don't resolve cleanly without needing deep liquidity or an oracle to verify the answers.

Prediction markets are great, but they have two bottlenecks.

1. Liquidity: Markets only get sharp when they attract enough participation, and that only reliably happens for near-term questions with fanatical bettors. Long-horizon, niche, complicated questions? They sit empty or thin.

2. Settlement: Most of what society actually argues about ("Are tariffs helping the economy?" "Is immigration too high?") doesn't have a clean outcome to resolve against. No oracle. No answer key. And without one, forecasts drift toward what's popular, not what's true. Truth is a weak equilibrium in the attention economy.

So we decided to tackle this problem from first principles. The mechanism is strangely elegant. Ask people what they believe and what they think others believe. Score the gap between what people predicted vs what was actually reported. It's called Bayesian Truth Serum (BTS), and over many rounds, the best strategy is boring yet powerful. Tell the truth and be well-calibrated.

We spent 2 years adapting BTS for continuous probabilities, hardening it against cartels, and wiring it to leverage @Starknet for execution and @worldcoin ID for identity verification.

The result is a protocol for collective human intelligence. Sharp crowd forecasts for questions prediction markets and traditional forecasting tournaments can't touch.

First competition starts in 7 days. If you're the type of person who sees things others haven't caught up to yet, this is your arena.

Register now👇

[Olena Protocol @OlenaProtocol]

Registrations are open for our Peer Prediction Competitions (link below).

Secured by Starknet (execution) + World ID (identity) 🤝

Our protocol incentivises honest reporting for any question - not just event-settled ones - and statistically aggregates inputs into a sharper crowd forecast.

First questions drop in 7 days. Monthly competitions. Participants share 1% of future token supply.

Register via World ID, receive voting power (VP), earn Olena Reward Tokens (ORT) by performing well.

Shout-out to @Nethermind for their work on the World–Starknet Bridge.

Happy forecasting. If you tell the truth and your truth is insightful, you’ll do well. That’s all you need to know!

[krisoshea.eth @krisos]

Very happy to share that I’ve been accepted into Cohort 7 of the @ethereum Protocol Fellowship, run by the @ethereumfndn's Protocol Support team.

I first came across Eth back in late 2017. At the time it was just something I found very interesting. A neutral, permissionless system that anyone could use or build on, not owned by any company, state, or institution.

In 2018 I really fell down the rabbit hole when DeFi started kicking off. I found the idea of an open, composable financial layer that anyone could build on top of really compelling. It felt like a completely new design space.

From there, Eth became my entry point into more complex topics like mechanism design, protocol economics, credible neutrality, and the challenge of coordinating large groups of people around shared infrastructure.

Over time this really shaped what I chose to study, what I chose to work on, and the kinds of problems I spend my time thinking about.

Most of my work so far has been at the app layer. The fellowship is a great opportunity to move closer to the bare metal. I'm looking forward to working on the protocol itself and learning from the people who are shaping what Eth becomes next.

I’m still narrowing down the exact project I will be working on, but currently looking into a Reth-based prototype around partial statefulness/state expiry, with broader interests in block production, censorship resistance, FOCIL, ePBS, and block-level access lists.

More broadly, I’m excited by this next phase of Eth protocol development: scaling, privacy, censorship resistance, credible neutrality, and the renewed focus on the values captured by CROPS.

Finally, a huge thank you to @C1aranMurray for introducing me to Eth back in 2017, to @post_polar_ for encouraging me to go for this, and to @joshdavislight and @TMIYChao for giving me the opportunity to take part!

Excited for the months ahead, and determined to contribute something meaningful to the protocol.

[Ciarán Murray @C1aranMurray]

Two bottlenecks:

1. Human review. @OlenaProtocol is being built to alleviate this by scaling human intelligence to meet AI output.

2. Human attention. Not really a solvable problem. It's a bit like the 1st law of thermodynamics: attention can only really be moved around.

[binji @binji_x]

> be zcash
> one of the most carefully built privacy systems in crypto
> cryptographers, auditors, the actual best people in the field
> then, a bug surfaces
> it was sitting in two lines of code the whole time
> “looks obvious in retrospect”
> it always does
> this is not a zcash problem but a problem for every piece of software ever written
> smart people write code, smart people review it, but bugs ship anyway
> enter the doom take
> ai is now insanely good at finding bugs
> faster than humans, at scale
> “if machines find every bug then nothing is safe and trustless anything is dead”
> mfw the thing everyone is scared of is the thing that saves us
> enter formal verification
> instead of writing code and testing it
> you write down exactly what it should do, in math
> then you PROVE it, with a proof a computer checks
> been around since the 1950s
> not new
> testing only checks the cases you thought of
> the bug lives in the case you didnt
> a proof covers every possible input at once
> all of them
> if any input misbehaves the proof just fails
> you cant ship broken and not know
> the zcash bug under formal verification is not a subtle thing someone has to spot but a hole in the proof
> proof doesnt complete
> you find out BEFORE anything ships
> not 4 hours into a thread on a saturday
> “ok so why doesnt everyone do this”
> used to be slow and brutally hard
> lived in aerospace + chip design + nuclear
> writing proofs by hand = serious expertise + serious time
> then ai showed up
> plot twist
> the exact skill that makes ai scary on offense
> reasoning through huge amounts of low level detail fast
> is the skill that makes formal verification finally cheap
> ai writes the code AND the proof
> humans keep the one job that matters: deciding what “correct” means
> crypto cares more than anyone
> normal software patches the bug and moves on
> crypto code holds the money
> crypto code IS the rules
> being wrong is public and usually permanent
> highest stakes, worst margin for error, perfect fit for proving
> SO
> while the timeline argues about whether ai killed security forever
> some people are building for the other outcome via ai enhanced formal verification
> ai-assisted formal verification proven onchain can securely house ALL public software, this is the S in CROPS !!

[Ciarán Murray @C1aranMurray]

I think the probability that this Zcash bug was exploited is low but the market is pricing in the reality that there are almost certainly many more critical bugs across the entire industry just waiting to be found by AI.

The beatings will continue until security is hardened.

[Luke Metro @luke_metro]

From a young age, I have always wanted to be the exit liquidity for shareholders of artificial intelligence companies

[krisoshea.eth @krisos]

@divyanshh_kalra @ethereum @linuxfoundation @inventnoveltyHQ Dust yourself off and try again. It’s a numbers game, keep trying and keep upskilling, you’ll make it.

[Rod Wong, MD @docrodwong]

a standing ovation for daraxonrasib at asco. over 40k oncologists, entrepreneurs, investors, and patient advocates together celebrating revmed's breakthru in the fight against pancreatic cancer. u never forget these moments. it's what innovation is all about.

[Farza 🇵🇰🇺🇸 @FarzaTV]

Watch me control my computer with just my voice. This is the future of operating systems.

No hands.

GPT-Realtime 2.0 is very, very underrated.

Demo:

[Dankrad Feist @dankrad]

Are passkeys actually like a hardware wallet built into your phone? I'm afraid not
(They are still pretty useful for quickly creating hot wallets)

https://t.co/bnSx5SEIS8

[zefram.eth @boredGenius]

Introducing TamaSwap, the first provably unhackable DEX.

- No-free-lunch theorem, machine-checked via Lean
- Onchain HTML interface, forever online
- No protocol fees, no-code deploy to any EVM chain

Built with Tama + Verity. AGPL license.

The platonic ideal of xy=k.

[Daniel Von Fange @danielvf]

I've submitted 3 small PR's to solidity that speed via-ir compilation up about 11%.

It's fun to learn about the insides of the compiler we all use, so let's tour each speed up. 1/7

[Beachhut @beachhutpr]

Managing the misinformation crisis..@C1aranMurray of @OlenaProtocol - an open platform where information is verified - told @WIRED's @daithaigilbert at @DubTechSummit why he believes, in the age of AI, we need infrastructure to support human verification of online content.#olena

[Thorsten Ball @thorstenball]

https://t.co/PV4OLiTZnu

[mrs kzg.eth née kassandra @kassandraETH]

I want to get a bit more public about the work we at the Kohaku Initiative inside the EF are doing

I notice there's hype but there's also confusion. Best way to clarify things is to speak candidly and openly about what I'm working on day-to-day

🧵time (bc i dont pay twitter $)

[Cat McGee @catmcgee]

I'm convinced that AI and crypto are perfect for each other in many ways

- verifiable compute
- verifiable training
- AI resistent sybil defence (proof of humanity etc)
- agent payments & micropayments
- decentralized training & GPU markets

So I went down the rabbit hole to see how I could

- train a model in a decentralized way
- while still keeping some weights private
- and zk proving that it was trained humanely/ethically

That last one ran into a whole lot of problems and led me down an even deeper rabbit hole until I arrived at mechanistic interpretability

From the outside, AI weights do not appear to correspond to how they act in any meaningful way

Mechanistic interpretability (mech interp) is about opening up neural networks and discovering why the weights do what they do

It basically says "don't just trust the black box of AI" which reminds me a lot of the philosophy of crypto

Anyway, Claude and I wrote a curriculum mainly for myself to help me understand mech interp. Thought I'd open it up for other software engineers

If you can't read AI papers or PyTorch but would like to, this is for you. It walks through 6 interpretability papers and replicates them in PyTorch. You can run them directly in the browser (using my huggingface space)

Check it out yourself and let me know what you think!
https://t.co/qXp5WyaS6G

[Georgios Konstantopoulos @gakonst]

Open Sourcing Centaur: Multiplayer, self-hosted, secure agents for Slack.

Centaur has been transforming how @paradigm and @tempo invest, build and research.

Now you can run it yourself on infrastructure you control. Instructions below.

[Ciarán Murray @C1aranMurray]

The problem for Ethereum sentiment is that the costs of its governance approach are visible but the benefits are counterfactual.

We can see comparatively slow execution, messy governance and weak coordination every day. We can’t see the corporate capture, regulatory capture and other governance failures that Ethereum’s structure helps avoid.

Or can't see it yet anyway. But stay tuned. As there's no way Hyperliquid, for example, will get to keep all the priviliges that come with being a decentralised blockchain when it isn't.

[krisoshea.eth @krisos]

@sodofi_ Didn’t realise you had joined the EF, congrats that’s huge! They’re really lucky to have you. EF will benefit immensely from havings devs like you who can create really polished content too. Looking forward to seeing the work you do with them!

[Ciarán Murray @C1aranMurray]

Charts like this make me less convinced that AI will destroy most human work, and more convinced that it will change where human work sits in the production chain. Yes AI clearly makes production cheap: there are already *many* more books, legal filings, scientific papers, code, analysis, claims. But a lot of this output still has to be quality controlled!

So the bottleneck moves from creation to review. AI can draft, summarise, structure and prepare work for inspection. In many cases, any situation where the cost of being wrong is big, it does not remove the need for humans. It simply changes the point at which humans are most valuable.

Instead of doing the whole task end-to-end, more human work becomes a targeted act of judgement: assess the likelihood a claim is correct, spot risk, sharpen a forecast, ensure a legal contract or software code reflects intention, or just generally decide whether a certain piece of info should be trusted.

So where I see things going is that AI makes a lot more professional work modular. By that I mean the unit of work gets smaller, clearer and more last-mile. The reason why is that AI does not just create the demand for human review. It also helps prepare the work for review. It can summarise context, structure the task, flag likely issues and bring something close to completion before a human sees it. That makes it possible to route shorter, better-scoped judgement tasks to the right person. The reviewer will heavily use AI of course too!

I think the inevitable outcome here is a need for a new kind of platform for facilitating human review: one that helps buyers find people with the right domain knowledge and helps reviewers find work where they can genuinely add value.

IMO the key to alleviating the bottleneck will be proper performance tracking. AI has benchmarks but human intelligence is generally measured through proxies. If we can change to something much more performance-derived, we can create a much more efficient market for human intelligence.

[erigon.eth @ErigonEth]

EVMone is one of the most optimised EVM implementations ever written.
It powers Erigon's execution engine and holds multiple performance records.

Zilkworm is built on top of it.

Every proving run starts from an execution trace generated by code that has been optimised for years at mainnet scale.
Not a re-implementation. Not an approximation. The actual engine.

What architecture choices matter most to you in a ZK prover?

https://t.co/jPKRgTUpQm

#Zilkworm #Ethereum