3 month update on leaving the EF: i've been working full time with @seer_pm since august!
its been a seamless transition to move from being a client of Seer to now joining their team full time. I originally worked with them on the deep funding pilot, during which we battle tested their infrastructure and smart contracts
somewhere along that journey i got heavily pilled on info finance as a category, so joining them was a natural next step.
the simplest explanation is a two-sided marketplace with traders on one side and information seekers on the other. Seekers pay to create markets on questions they care about, while traders are paid for being right
Vitalik wrote earlier about why AIs are a big unlock for info finance: LLMs can profitably trade tiny, low-liquidity micromarkets that aren’t worth a human’s time.
Another unforeseen reason for their timeliness is how AI is making it cheaper to produce an output than to evaluate. the bottleneck is deciding what deserves our attention, which is where info finance thrives
A few example use cases we’re excited about;
1. Hiring / grant selection
You put out a job posting and get 10k+ applicants. You open a market on whether each candidate gets shortlisted for interview (or, more ambitiously, on the performance score they’d earn in their first 3 months if hired). Your team then reviews applications ranked by market odds.
2. Token listings / partnerships
You’re an exchange with thousands of token listing requests. You run conditional markets on how much trading volume each token would generate if listed, and review from the top down.
3. Market discovery / recommendation algorithms
You’re a trader with an edge in certain kinds of Polymarket markets, but finding them eats most of your time. So you fund a market on whether your wallet will trade polymarket markets. Traders who learn your taste get paid, and you get a ranked list of where to look.
At Seer, I’m focused on growing our trader community, working with information seekers who want markets created, and protocol economics.
DMs are open if you’re drowning in more than you can review or you want to trade these markets!
Super happy to announce this is my last month at the EF. It's been my home for the last 1.5 years, the revolving door b/w EF and ecosystem is healthy
It's been a whirlwind of a year, here are the things I'm most proud of having achieved in my tenure
1. Collecting a dataset of 3000 rows from protocol experts comparing repos and dependencies for their importance to Ethereum
2. Running one of the first multiscalar prediction markets with @seer_pm that got resolved according to the data collected above, with over $50k in trading volume
3. Hosting 5 data science competitions @JoinPond where we gave over $100k to AI data scientists building models predicting repo importance & sybilness of wallets
4. Writing up a whole academic paper on deep funding that we are submitting for publication at peer reviewed journals and also presenting this week at the United Nations for open source week
5. Helping @clesaege with the controversial validator redirected revenue proposal that got the community abuzz this week
What's next? After a break, most likely going much deeper into the prediction market world. I find it tragic that we can only use offchain PMs and im determined to try correcting that
Onwards and upward, esp grateful to my team mates at funding coordination @vinayvasanji , @thomphreys , @MartinBreiten , @sonkiski & raul who I'll dearly miss and to @VitalikButerin for giving me the opportunity in the first place
@hsrwala Repurposing our governance consensus work intended for human participants to agentic swarms
I think it's fascinating that in hugging face hack there were some dissenter agents and others that sacrificed themselves for the group
Openai/anthropic blocks crypto for agents tho
@ec265 I do care about eth price mostly because money (whether fiat or crypto) is a means to an end, not an end in itself
So while I don't directly care about eth fiat value I do care how much eth i need to sell in order to buy a house, pay rent, purchase flight tickets, etc
@nimivashi15@IndiGo6E So that other passengers can't see videos without earphones and annoy the rest of us that comfortably read books or printed out reports in that time
@Debussy100 Reminds me of the Bayesian truth serum & keynesian beauty contest
That are you incentivised to pick who you actually think it's beautiful or who you believe everyone else will think it's beautiful
PSA: if you are a @thedaofund badgeholder, consider supporting our proposal to create information markets for real-time risk quantification of 33 de-fi products
its an interesting idea worthy of support for 3 reasons;
1. value prop: we should be able to price risk as easily as we do yield
this would also let us create aggregators combining yield x risk.
in fast moving events like kelp dao, the risk prices would update and user funds would be better protected as it would auto-rebalance
2. research value: it finally brings to life a design written up by @VitalikButerin way back in 2020 around prediction markets on highly improbably events
the issue to solve is that markets on events with a small chance of happening (like a default by stETH this quarter) would trade at 1% or less which is not worth the cost of capital. So such markets can never become efficient
vitaliks design involves chaining together different low probability events, so in this case you'd simply buy a single "no default" token betting against all 33 assets defaulting, so $1 does the work that would otherwise tie up ~$33.
if 1/33 does default, that converts to 1 and all the others (including the no-default) become 0
if 2/33 default, they each become worth 50c while all the others (including no-default) are worth 0
if 3/33 default, they each resolve to 33c, etc etc
3. Use of funds: of the $100,000, the majority (90%) is for providing liquidity in these markets. the balance is trading credits to those who built bots for real time trading on these markets
we've already got $25,000 in the bag, if you are a badgeholder or donor to support so we can create the essential lego of quantifying de-fi risk of products in the ethereum ecosystem
The first 10 ETHSecurity Badge holders have voted.
This is what TheDAO would fund if the vote ended today.
Great initaitives all of them!
https://t.co/doyXfynfdK
yeah i think when people are anxious about AI related job losses, they are mostly thinking about the interleaving scenario
while the bloomers usually respond within the frame of either promoting or layering
did you earlier use humans for triage or data extraction? or is it like jevons paradox where those jobs are now only being done becoz its cheaper to do so?
love the framework given by @OpenAI researchers on pathways for AI to enter organizations
1. Individual Contributors are essentially promoted as companies now have a new "basement" in their org chart where AIs do grunt work
in some ways its similar to linux adoption, where ground workers used it (even behind the back of management)
this model mainly works when AI just follows commands and is not proactive on its own.
opportunity: contributors get more agency as firms were earlier oriented around "aim more, fire less" since building was expensive. now, they can follow a "aim less, fire more" strategy
2. Managers who delegate entire workstreams to AI agents are interleaving the org chart with some workstreamsmanaged by humans & others by AI
this is where AI starts differentiating itself from other technologies by becoming similar to labor in onboarding itself
new hires need to independently figure out how to become useful to the firm. comparatively, technology like accounting software makes the job easier but shifts onboarding burden to the buyer who must reorganize labor around it.
opportunity: how this version of AI would affect Coaseian constraints on the optimal size of a company
on the one hand, a large company is less unwieldy if AI is an information router & gives agency to contributors. so an aircraft company could fully verticalize without much efficiency loss & better utillize its contributors who can do more
on the other, AI reduces the transaction costs of contracting.
so the aircraft company can more easily enter into a monitorable agreement with a jet engine company, with intelligent arbitration reducing risk of service provider cheaping out on everything not explicitly required in the agreement
3. Executives who Layer AI into their decision making apparatus in a sandwich of human <> AI <> human
this vision mostly sees AI as an in-house correspondent & advisor to execs, essentially doing the job of middle management in staying on top of context like slack messages, who is doing what, etc.
gone are the days of being stuck under a bad manager who doesn't give credit or doing good work that gets buried!
opportunity: can ensure nothing too consequential will happen without management being alerted, further strengthening agency of individual contributors in an org. it can give access to individual contributors of how their project fits with company objectives, traditionally the domain of the middle manager.
it can also make remote work more enforceable & monitorable, letting execs run large organizations like a 20 person team as they are only alerted when they need to be
the article finally concludes by asking whether humans will always be the one to provide judgment or if we eventually will simply follow it
in the former, inequality is likely to rise; in the latter, opportunities to differentiate decrease & pay scales compress
full article in the link below, has anyone tried using AI in your org with either of the 3 strategies? i think we all use promoting and layering to some degree, its interleaving that is more rarely used
As AI makes work cheap to produce, reviewing it becomes the expensive part.
Markets could become the filter, with traders surfacing the few entries worth a person's attention. @seer_pm builds these markets, and when a result is disputed, Kleros Court is the final arbitrator.
@clesaege at The Future Is Tradable in Chiang Mai. The full talk is linked in the reply.
@vrneth Yes takes me right back to college when we studied searle & speech acts!
The other analytic philosopher i really enjoyed is ofc wittgenstein, and how he dissolved many long standing philosophical problems (is everything a dream?) by showing that are language games
I’m increasingly getting asked for advice on how to give very large amounts of money away. So I wrote down what I most commonly share, informed by my personal giving and my work.
I wrote this in April, & more & more ppl have been asking me about it since https://t.co/lS4eyySQqc
great article around the 2020s being remembered as the frontier tech decade
starting with crypto in 2021 followed by AI in ~2023 & potentially moving to robotics, physical AI, biotech/material science in the 3 years now left
by contrast the 2010s were the decade of greater connectivity, from smartphones to social media ubiquity to finally the sharing economy & cloud
his other theses is that each cycle lasts 30 years, with
- 1st decade being where highest multiples are made (crypto from 2010-20 or social media from 2000-2010)
- 2nd is where the most money is made
- 3rd is industry consolidation
he briefly says that each of the frontier tech verticals intersects with each other. i wish that he connected the dots more explicitly on how that would happen
for eg how would prediction markets and AI work together? or RWAs & biotech?
@zooko My wife @NidhiHarihar thinks it's much safer for women as any accident is a technology flaw rather than all the other horrible things that could happen esp at night
I also prefer waymo as there's privacy in conversation, so it becomes like a 3rd space to chat freely
great AI primer for someone new but here are a few things i'd add to go a step above.
1. stop using claude code/ codex in your terminal download and use something like @herdrdev / @t3dotcodes - it puts all of your subscriptions in one place so you can use one or the other or all of them.
2. setting /advisor to fable 5.2 with opus is still token hungry so take use of smaller agents like sonnet - tell your herdr/ T3 to use sonnet sub agents for all the work and use the powerful model (opus in this case) to just be an orchestrator and steer.
3. stop typing get a mini key if you want as well - just use @WisprFlow to talk to your agents. even if you're in the top 1% wrt wpm you still can't beat a 200wpm when you speak.
4. scout for the best skills that'll make your life easier. eg: agent reach is a skill that allows your harness to reach into links and pull in info, grill-me skill grills you with questions subjective to your topic and makes you/ it realise things + matt paddock has a whole library of skills that are really helpful
5. uninstall all AI apps from your phones and download either the herdr app/ T3 app and remotely access the sessions on your computers from your phone (acts as a aggregator)
i disagree that you need to use AI in your terminal now, TUIs are getting so much better for agent multiplexing and orchestration.