Delphi Digital

A research-driven firm dedicated to making crypto happen sooner and better than it would without us.

[Delphi Digital @Delphi_Digital]

Our New Report "Hermes: The Moat Above the Model" is live!

Hermes is the AI agent that belongs to you.

Your agent builds valuable context about you with every session. When you switch models, you can't take any of this with you. What a closed agent learns about you is its product. You are renting your own work back.

Hermes breaks this model by keeping everything it learns on your machine. Between sessions the agent reviews its successful runs and writes them into reusable skills using GEPA, which beats reinforcement learning with 35x fewer rollouts. These skills travel with you when you switch models.

Two years before Hermes shipped, @NousResearch was already building the fine tuning and reinforcement learning pipelines for their open-weight models. In the last week of May, Hermes was OpenRouter's highest volume app with 4.5 trillion tokens routed currently.

At Hermes's scale, that signal lets Nous keep improving the user experience. Your memory and skills stay on your machine the whole time.

[Delphi Digital @Delphi_Digital]

Our New Report "Hermes: The Moat Above the Model" is live!

Hermes is the AI agent that belongs to you.

Your agent builds valuable context about you with every session. When you switch models, you can't take any of this with you. What a closed agent learns about you is its product. You are renting your own work back.

Hermes breaks this model by keeping everything it learns on your machine. Between sessions the agent reviews its successful runs and writes them into reusable skills using GEPA, which beats reinforcement learning with 35x fewer rollouts. These skills travel with you when you switch models.

Two years before Hermes shipped, @NousResearch was already building the fine tuning and reinforcement learning pipelines for their open-weight models. In the last week of May, Hermes was OpenRouter's highest volume app with 4.5 trillion tokens routed currently.

At Hermes's scale, that signal lets Nous keep improving the user experience. Your memory and skills stay on your machine the whole time.

[Delphi Digital @Delphi_Digital]

Watch the full episode here.
https://t.co/CERyYtfAgI

[Delphi Digital @Delphi_Digital]

Kevin explains why Hyperliquid's outperformance is important for crypto.

"It’s good for HYPE to be outperforming right now... because it’s arguably one of the most important potential catalysts to drive more capital into this market.”

[Delphi Digital @Delphi_Digital]

Read the full report here. First 20 people get it for free.
https://t.co/yL7480rsQA

[Delphi Digital @Delphi_Digital]

Our New Report "Hermes: The Moat Above the Model" is live!

Hermes is the AI agent that belongs to you.

Your agent builds valuable context about you with every session. When you switch models, you can't take any of this with you. What a closed agent learns about you is its product. You are renting your own work back.

Hermes breaks this model by keeping everything it learns on your machine. Between sessions the agent reviews its successful runs and writes them into reusable skills using GEPA, which beats reinforcement learning with 35x fewer rollouts. These skills travel with you when you switch models.

Two years before Hermes shipped, @NousResearch was already building the fine tuning and reinforcement learning pipelines for their open-weight models. In the last week of May, Hermes was OpenRouter's highest volume app with 4.5 trillion tokens routed currently.

At Hermes's scale, that signal lets Nous keep improving the user experience. Your memory and skills stay on your machine the whole time.

[Delphi Digital @Delphi_Digital]

LIVE NOW: Delphi's research team is walking through State of Token Markets, the data behind why this cycle failed altcoin holders. Plus a live demo of the Token Design Toolkit.

Join @0xdaft and @ashwathbk -> right on our profile page.

[Delphi Digital @Delphi_Digital]

Walking through the Token Report and Tool https://t.co/CBOeQA00O2

[Delphi Digital @Delphi_Digital]

Watch the full episode here.
https://t.co/CERyYtf2ra

[Delphi Digital @Delphi_Digital]

A new episode of the Hivemind is live!

This week we discuss Bitcoin's breakdown, Hyperliquid's future, Crypto fundamentals, and more.

Timestamps:

0:00 Introduction & Is Bitcoin Cooked?
5:44 Saylor, Bitcoin Liquidity & Market Structure
12:28 Why Crypto Is Finally Detaching From Bitcoin
19:06 The New Altcoin Market & Capital Rotation
22:35 Why HYPE's Outperformance Matters
26:22 Zcash, Value Investing & Crypto Fundamentals
31:57 Hyperliquid, Lighter & Crypto's First Compounders
33:47 Can Crypto Thrive While Bitcoin Stagnates?
38:32 Why Retail Left Crypto For AI & Stocks
40:46 BlackBerry, Equities & Staying In Your Circle Of Competence
47:54 Lighter Deep Dive
53:16 Kalshi, Perpetual Futures & Regulatory Tailwinds
55:03 Polymarket's UMA Problem
1:07:09 Ethena x Coinbase Explained
1:16:43 Bitcoin To $49K? Jason's Bear Case
1:19:43 Can HYPE Keep Going Higher?

[Anil Lulla @anildelphi]

Our Consulting team @Delphi_Digital works with just a dozen projects a year, becoming extensions of their team to help with everything from token design to gtm to product development

They put together this report to give away a lot of the insights they’ve gathered from the work they’ve done

In it, they break down exactly how and why token markets broke and dissect what works and needs to change to fix them

Instead of setting this behind a paywall, we wanted to make sure anyone who was interested in reading it could access it for free. Link below to read the whole report.

[Delphi Digital @Delphi_Digital]

Token markets are broken by design.

Buying every new listing across the major CEXs since 2025 would have lost retail about half its capital.

The median listing fell 82%, and exchanges have become a place where early holders sell into retail buyers.

Tokens with monthly unlock schedules underperformed BTC by about 7% in the week leading upto and weeks after each unlock event.

What makes token outperformance difficult is that the next unlock usually arrived before the market has had the opportunity to absorbed the last.

Six major airdrops saw 78 to 94% of recipient wallets sell off most of their allocation by day 90. One of those programs paid roughly $1 billion in token value to wallets that haven't stayed around.

The best designs are simple. They now route revenue back to holders and tie supply releases to performance milestones. A revenue-weighted basket of the 10 highest-revenue protocols has outperformed BTC, ETH, and SOL since January 2025.

What's left is one of the strongest setups crypto investing has had: fewer tokens, better designed, and structural buyers replacing the hedge funds that exited.

[neel daftary @0xdaft]

https://t.co/PnuiNORwSj

[Delphi Digital @Delphi_Digital]

Watch the full episode here.
https://t.co/8cGQUhtniz

[Delphi Digital @Delphi_Digital]

Jason explains why Bitcoin is struggling to find a bid.

“All of the buyers and the people responsible for driving Bitcoin up to and through $100k are gone. Many of them are underwater and a bunch are selling.”

[Delphi Digital @Delphi_Digital]

Airdrops as we know them are over.

Giving tokens away to build a holder base has mostly created sellers. Across the largest airdrops, between 78% and 94% of recipient wallets had sold off most of their allocation by day 90.

People cite Hyperliquid and Jito as proof airdrops can work, but neither succeeded because of the airdrop. Hyperliquid had over $1B in revenue funding buybacks that soaked up the selling from recipients. Jito’s eligible cohort was small enough to avoid industrial farming.

Token economics are starting to require real protocol performance. MegaETH locked 53% of its supply behind performance targets. Pendle routes roughly 80% of revenue into buybacks for stakers.

Token distribution is moving from handouts to performance.

[Delphi Digital @Delphi_Digital]

Watch the full episode here.
https://t.co/8cGQUhtniz

[Delphi Digital @Delphi_Digital]

A new episode of Market Matters is live.

AI continues to lead equities higher while crypto struggles to find a bid.

Our Markets team discusses the AI rally, Bitcoin’s underperformance, the real-yield backdrop, and more.

[Delphi Digital @Delphi_Digital]

Join us tomorrow for a live walkthrough of State of Token Markets, Delphi Consulting's newest report.

1PM ET, streaming live right on here.

We'll cover: Why 46% of 2025 CEX listings lost more than 80%. Airdrop math. Unlock death spirals. What HYPE got right. Plus a live demo of the Token Design Tool.

With authors @0xdaft and @ashwathbk.

[𝕋𝕖𝕞𝕞𝕪🦇🔊 @Only1temmy]

DELPHI DIGITAL JUST DROPPED THEIR 2026 STATE OF THE TOKEN MARKET

HERE'S THE TLDR:

1/ 2024-25 had every ingredient of a bull market. But for altcoin holders it was one of the most dilutive periods in crypto history. Institutional money went to BTC and ETH. Everything else got left behind.

2/ Of tens of millions of tokens minted this cycle, only ~1,700 (under 0.01%) generate $250K+ in daily volume. The long tail is basically a graveyard.

3/ Buying every CEX listing from Jan 2025 destroyed $329K of every $652K invested. Median return: -82%. Gateio alone listed 348 tokens, more than the other four exchanges combined, with an 85% loss rate. Volume without curation was the whole business model.

4/ Most tokens spend their entire lifetime below their launch price. L2s are the worst, 80% of lifetime underwater. Consumer crypto is a lottery ticket. CEX tokens are the only cohort that actually got better over time.

5/ For most tokens in 2025, doing nothing beat buying and holding. Only tokens with real structural buyers on the other side (HYPE, SKY, ZEC, XMR) consistently held. Everything else just bled.

6/ Most VC-backed tokens launched this cycle are below their launch price. The worst are down 90%+. Inflated FDV, low float, brief pop for insiders and airdrop farmers, then relentless sell pressure as unlocks kick in. Holders have no revenue claim, no floor. Death by a thousand cuts.

7/ $TRUMP launched Jan 17, 2025. Peaked +576%. Collapsed. Took the whole memecoin category with it and the liquidity never came back. The Memecoin Index is down 75% since. 64 weekly rebalances across dozens of tickers couldn't fix it.

8/ Only 6 memecoin names held continuously since Jan 2025: BONK, WIF, FARTCOIN, FLOKI, PEPE, SPX. Everyone else is a round-tripper or dead.

9/ Launch FDVs fell 85% from Q1 2024 to Q1 2026. Still wasn't enough. Tokens lost another 30-80% in the 6 months after TGE anyway. The market reprices faster than teams can launch.

10/ Unlocks don't just hurt, they compound. Each event erases ~7% excess return vs BTC in the surrounding 3-week window. 28 of 33 tokens underperform BTC at every single unlock. Worst offenders: PEAQ -19.8%, VENOM -18.6%, W -15.6%, ZK -15.1%, ANIME -12.0%. The problem isn't the unlock. It's that the next one is 30 days away.

11/ Launchpads vs CEX: at 7 days every launchpad cohort is green, every CEX cohort is already underwater. The one exception is MetaDAO, where Futarchy governance let holders vote to unwind and receive $0.822 back on a $0.80 sale. A floor no CEX listing has ever provided.

12/ Airdrops are structurally dead. Sybil cost is now zero in an agentic world. Legitimate issuers can't legally do them. The CAC math is catastrophic (ARB alone paid ~$1.36B to users who left within a month). The ones that worked (Hyperliquid, Jito) were coincidences, not templates.

13/ Token holders have no rights when it matters. Pumpdotfun, SOL Strategies, Circle, Coinbase all did acquisitions this cycle. In every case equity captured the value. Token holders got nothing or close to it. Until token rights are legally enforced, holders rely entirely on team goodwill.

14/ The DAT premium is gone. Of $104B across 59 pure-play DATs, Strategy alone holds $63B. The other 58 trade at a median 0.81x mNAV and 36 of 59 trade below the actual value of their crypto holdings. The staking yield edge over ETFs is gone. MARA is already selling BTC. The deleveraging cycle has started. Strategy was the only real winner of that model.

15/ Revenue-weighted portfolios returned +30.6% since Jan 2025. BTC -17%, ETH -35%, SOL -58% in the same period. Cash flows beat narratives. It's not even close anymore.

16/ Fundamentals were always priced in. The market just took two years to prove it. CEX tokens, the only category backed by issuers with measurable recurring revenue, trade at 8.9x the broader market at the 2-year mark. L2s collapsed to 0.16x. The data sorted what narrative couldn't.

17/ The fee switch era is here. Between late 2024 and early 2026, every major DeFi protocol with real revenue either launched with value accrual or voted it in. HYPE, SKY, AAVE, JUP, PUMP, MET, UNI, PENDLE all flipped. "The fee switch didn't arrive. Teams stopped fighting it."

18/ But buybacks alone don't save a token. HYPE +533%. SKY +25%. AAVE -25%. JUP -40%, despite having the highest buyback yield in the set at 18.83%. Token-specific pressures can overwhelm any buyback program. "Buybacks are the mechanism. They are not the moat."

19/ The math shows why. Aave barely covers its own emissions (0.90x coverage). Jupiter is overwhelmed: $3.70 in unlocks for every $1 bought back (0.26x coverage). You can't buyback your way out of a broken token structure.

20/ The ETF holder base quietly rotated. Brevan Howard -27%, Tudor -72%, Hunting Hill -97%. All basis-trade unwinds. Who replaced them: Morgan Stanley +177%, BlackRock proprietary (brand new position, built to 51.4M shares), Mubadala/Harvard +54%. Advisors up 204%, sovereigns and endowments up 228%. The leverage trade left. Real long-term capital arrived.

21/ Emissions must become a function of performance, not time. Five mechanisms are emerging: performance-gated unlocks, retroactive supply destruction, supply-cut governance, fair-launch structures, and liquidity-adjusted vesting. Before 2025 the trigger was a calendar. Going forward it should be earnings.

22/ The protocol scorecard is brutal. HYPE passes on everything. AAVE passes on nearly everything with a partial on value accrual. JUP and PUMP fail on emissions. PLUME and BEAR fail on product, revenue, and emissions. TAO has narrative but no revenue. Most tokens have not earned the right to exist as investment assets yet.

23/ The market stopped buying stories. KAITO -76%. PLUME -94%. BERA -94%. LINEA -87%. All had great narratives, none had traction. HYPE, AAVE, UNI won because the numbers were real. "A clean story still helps. It just isn't enough."

24/ Memecoins that pretended to be revenue protocols got wiped. AI16Z -95%, ZEREBRO -96%, GORK -98%, ACT -98%. The survivors (DOGE, SHIB, PEPE, WIF/BONK, FARTCOIN) are the ones that were always honest about what they are. Tradeable attention. "The most interesting thing that can happen to memecoins is for them to become honest."

25/ The whole 2021-2024 era was crony economics. VC cap tables, governance-only tokens, calendar dumps, pay-to-play listings, retail as exit liquidity. All of it is being replaced. Fewer tokens, real revenue, holders with actual rights, and buyers who don't exit in 30 days. That setup has never existed in crypto before. Now it does.

[nate @akachacolate]

the power of data, great read

[Delphi Digital @Delphi_Digital]

Read the full report for free here.
https://t.co/YeBITnqGwv

[Delphi Digital @Delphi_Digital]

Airdrops as we know them are over.

Giving tokens away to build a holder base has mostly created sellers. Across the largest airdrops, between 78% and 94% of recipient wallets had sold off most of their allocation by day 90.

People cite Hyperliquid and Jito as proof airdrops can work, but neither succeeded because of the airdrop. Hyperliquid had over $1B in revenue funding buybacks that soaked up the selling from recipients. Jito’s eligible cohort was small enough to avoid industrial farming.

Token economics are starting to require real protocol performance. MegaETH locked 53% of its supply behind performance targets. Pendle routes roughly 80% of revenue into buybacks for stakers.

Token distribution is moving from handouts to performance.