
research & growth @Shoalresearch
New York, NY
Joined September 2021
- Tweets4.9K
- Following2.6K
- Followers4.4K
- Likes10.5K
Pinned Tweet
things i’d like to see in 2026
(in no specific order)
1. the privatization of defi
- infra and apps that integrate privacy-preserving features for end-users
- deployments have already started with swaps, and I expect they will work up the rest of the defi stack (perps, borrow/lend, LP, etc)
- most will converge around these two designs:
a) plug-in solutions: chain-agnostic protocols that integrate on top of existing chains and apps
(e.g. @zama @UmbraPrivacy @encifherio)
b) new chains: new L1s and L2s with natively embedded privacy
(e.g. @octra @0xMiden)
2. internet-native payments
- embedding payment mechanisms into the fabric of the internet is not a new idea by any means, but we've never had the amount of infra and tooling to get this right as we do now.
- I expect we'll see the first major growth spurt in developer activity around x402
- I also expect multichain/interop efforts surrounding x402 to become more critical as new stablechains like @tempo and @arc go live
3. stonks onchain
- crypto unlocks global retail access to the next major technological sectors (AI, robotics, quantum, etc)
- more collaborative efforts between crypto builders and regulators like the SEC and DTCC
- 25x lev perps on SpaceX soon?
- the demand for microcaps and shitters will flow into onchain stocks instead
4. fixing the on/offramp problem for stablecoins
- stablecoin pmf is clear
- but there is still lots of room for improvement
- on/offramping is one of if not the biggest pain points remaining
- conversion fees start to trend down as we see more market entrants join
5. new prediction market primitives
- we'll see more and more teams seeking to differentiate and build unique products and services in the hottest new sector in crypto
- lots of focus on yield (money markets, LP)
- prediction market perps roll out
- could see lots of demand for private prediction markets; I expect someone will launch this in Q1
- and with all the prediction market arb bot threads coming on the timeline recently , I could also see demand for more compliant proof-of-personhood based prediction markets (maybe someone launches this on @worldnetwork)
- which brings me to my next point
6. legally compliant privacy
- products that allow users to verify their identity in a legally compliant manner while obscuring sensitive data (e.g. @onflowxyz)
- can see lots of demand for this at scale
- adoption skyrockets if a major wallet or exchange integrates this as an auth method
6. AI-powered finance proliferates
- a meaningful chunk of financial activity in crypto will be generated by AI agents, executed on behalf of their human overlords
- anticipate lots of demand and growth for teams like @gizatechxyz
- we’ll see several teams launching agentic frameworks (packaged through accessible interfaces) for people to build custom onchain strategies with (i.e. @Cod3xOrg Send AI)
- and it won’t just be onchain or crypto-native companies pushing this, I don’t see why fintech giants like Robinhood wouldn’t want to launch their own prototypes
7. more ICOs, and more innovation around launchpads
- look 2025 was undisputably rough for new token launches.
- that said with the regulatory shift in US, i believe there will still be plenty of appetite and demand for ICOs in 2026
- with the renewed demand, we'll see new products and services being built around ICOs as well
- i'm particularly interested in private/confidential ICOs
if you're building on any of these frontiers, my DMs are open. 2026 will be a big year.
Paul Timofeev retweeted
Replying to @birdabo
Really good take
Reading and consuming good things -> producing better outputs (e.g. writing)
AI helps read and consume much faster but also much less content overall, effectively starving the source for good outputs
-> Read more
AI has made my research better. But I realized today it’s also costing me something...
When I was in middle school, I asked my grandpa what a word meant. He told me to go grab the dictionary and look it up. I asked why I couldn’t just look it up online. He said that would get me the answer for sure, but I’d be missing something. When you open a dictionary, you see all the words next to the one you’re looking up, and you end up learning a few of those too. Looking it up online gets you the one word and nothing around it. No serendipity.
Zooming out, the internet era already lost some of that serendipity. It used to be that for a research project, you had to go find a book in the library. On the way to that book, you’d see books you’d never heard of. And while you were looking through its table of contents for the information you needed, you’d inevitably pick up other concepts too.
Internet research is much more directed. What you’re trading is access to more information, faster, but with less variance. Still, the internet was unique in that you could chase down links and go down rabbit holes. There was still some serendipity to it. It wasn’t a full departure from the past.
AI research pushes that trade even further. It often one-shots the research, the same way looking up a word online does. You don’t need to search around, so you rarely stumble onto adjacent ideas. And a lot of the most interesting ideas come from hearing about them by accident.
To be clear, I’m an optimist here. AI research has made me a lot better and lets me take in more information than I ever have in my life. But for myself, I’m trying to keep a serendipity mindset and expose myself to enough content that I catch things I’d never have expected.
Paul Timofeev retweeted
Medal Declined $500M From OpenAI: Game Clipping to General Intuition's Spatial Intelligence
Shoal Signal Ep. 20 with Ken Colton, CEO at @Medal_TV, hosted by @zaddycoin
We cover the $500M offer and why Medal spun out @gen_intuition instead, six million gamers clipping every month, spatial training data, why player judgment cannot be pulled out of simulation, action policies versus world models, and why @ken_colton thinks shared gaming moments will soon outnumber IRL ones.
0:00 Intro and Ken's background
1:28 6,000 hours of Rocket League
3:22 Why Rocket League hooked Ken
4:48 Mods, tools, and the path to Medal
10:00 Medal versus General Intuition
17:38 OpenAI shows up for the data
18:45 Deciding to build it themselves
20:32 Six million gamers a month
22:21 Turning controller inputs into actions
28:30 Training on player intent, not game engines
30:46 Action policies versus world models
34:48 Why simulation misses player judgment
41:19 Gaming data is easier to gather
42:26 The multiplayer world model
47:49 Game bots first, robotics later
51:01 Gaming moments will outnumber IRL moments
55:01 Not TikTok for gaming
58:24 Ken's closing alpha
Someone's gotta build the picks and shovels
arc alpha
@arcpad_meme - launchpad
@tollylabs - launchpad
@circlewarp - launchpad
@minarafun - launchpad
@ArcDEXScan - launchpad
@Archemistdotfun - launchpad
@flutchdotfun - launchpad
@ubidotfun - launchpad
@actfunxyz - launchpad
@akadotfun - launchpad / dn404
@Arcanedotfi - launchpad
@Zyoradotfun - launchpad
@rwarcdotfun - launchpad
@sashimidotfun - launchpad
@eve_dot_fun - launchpad
@Ayooclub - launchpad
@liftdotfun - launchpad
@focidotfamily - launchpad
@synthra_finance - launchpad / amm
@Orixaxyz - marketplace / launchpad
@chaos_launchpad - launchpad
find the first runner on arc mainnet, it will reach hundreds of millions
if you got any insiders let me know in my dm, thanks in advance
do your own research before clicking any projects
Paul Timofeev retweeted
I have been privileged to spend the better part of the last decade building my life here in NY - an opportunity that was stolen from nearly 3,000 people on September 11, 2001.
The world changed that day.
We must never forget it.
Paul Timofeev retweeted
Great read, covers how cloud players are thinking about the changing economics of the internet and x402.
The clearest explanation of how AI actually disrupts the open web as we know it today, and where we go from here
Grab a coffee and enjoy the read 🐚
Paul Timofeev retweeted
Managing Chaos in Atoms and Markets: Michael Egorov on Physics, Liquidations, and Financial Engineering
Shoal Signal Ep. 19 with @newmichwill, founder of @CurveFinance and @yieldbasis, hosted by @zaddycoin
We cover how cooling atoms to 100 nanokelvins taught him to trust an invariant instead of a price, why he took impermanent loss on a trade on purpose, the half year he sat in liquidation on purpose to test Curve, and why nuclear fusion gets solved sooner than the field expects.
0:00 Photons look like giant cucumbers
0:58 What coherence time actually is
4:46 Bose-Einstein condensate and laser cooling
8:47 A qubit that held for seconds
11:36 From ideal atoms to ideal markets
13:50 The invariant is the only signal
15:08 Control systems in the lab and DeFi
16:42 Quantum finance and prices tunneling
18:49 A toy universe made of vortices
21:18 Why there is no antimatter around us
23:57 Crypto as the revolution in money
25:29 Cycles from Mt. Gox to meme coins
30:40 Where retail's next opportunity sits
33:36 The Web3 promise that never landed
35:28 Why crypto fits agents better than humans
39:09 Working only on impossible problems
40:00 Solving impermanent loss with Yield Basis
40:48 Reversible liquidation in Curve
42:26 Half a year in liquidation on purpose
45:42 Where the risk actually sits
49:45 Paying authors when AI trains on books
53:28 Energy as the next breakthrough
57:17 Two roads to fusion
1:03:33 Thorium and the economics of nuclear
1:08:13 Where the alpha is
1:10:21 Turning mercury into gold in San Francisco
Paul Timofeev retweeted
I did it with Grok Bot. You gotta try Instinct. Send it to your Townie.Put it in Work mode. Give it to Cowork. It’s just Hatch. Tandem can do it. Give it to Juno. Nudge it to Pip. Dude just use Manus. I got a Duggie invite. OpenClaw on a Mac mini can handle it. Crundle is goated.





























