The Fed is on hold and explicitly warning you not to front‑run; BTC keeps failing near $70k and ETF flows aren’t a reliable bid. This week is, again, about optionality. Cash remains a valid position, convex hedges are welcome, and tight triggers are a must—until price and flows prove the next leg.
Key Takeaways
- Inflation re‑accelerated. Headline PCE +0.4% m/m (vs +0.2% prior), +2.9% y/y. Core (ex‑food/energy) +0.4% m/m (vs +0.2% prior), +3.0% y/y.
- The U.S. put a 10–15 day clock on an Iran nuclear deal (read: March 1–6 window) while a sizable military buildup sits in the background.
- BTC is the line: repeated failures near $70k = market still de‑risking. Treat $70k as “prove it,” $66–67k as the first dip zone, ~$65k as the hard line.
- ETH/SOL are conditional bets: ETH earns size only if it holds $2,050–2,100 with improving ETF flows; SOL is “buy dips” only while $80–83 holds.
- For alts, the leverage profile is already cleaner, but squeezes need fresh spot (not just “OI is light”). Don’t confuse positioning reset with a bid.
- Move too early and you burn capital building infra before the market is ready; move too late and you’re defending yesterday’s margins.
- Don’t run leverage into a potential credit air-pocket; cash/stables = optionality; buy convexity, don’t short.
Macro: Fed On Hold
January CPI came in at +0.2% m/m and +2.4% y/y, with core +0.3% m/m and +2.5% y/y; energy was a drag (-1.5% m/m) while shelter was still the main monthly contributor. On labor, January payrolls were +130k and unemployment held at 4.3% (unemployed 7.4M), with long‑term unemployed at 1.8M (25% of unemployed). That context preceded this week’s FOMC minutes, with the next FOMC meeting on Mar 17–18. The Fed is explicitly telling you not to front‑run an easy‑cuts glidepath: the January FOMC minutes show most officials wanted to keep rates steady (3.5%–3.75%) and warned progress back to 2% inflation could be “slower and more uneven,” with some even flagging that hikes could come back on the table if inflation re‑sticks.

Inflation re‑accelerated. Headline PCE +0.4% m/m (vs +0.2% prior), +2.9% y/y. Core (ex‑food/energy) +0.4% m/m (vs +0.2% prior), +3.0% y/y. Demand isn’t collapsing, but it’s not ripping either: nominal PCE spending +0.4% m/m, real PCE +0.1% m/m; personal income +0.3% m/m; saving rate 3.6%.

Zooming out, Ray Dalio frames the current scenario as a “breaking world order” where debt + internal politics + geopolitics increase the odds of policy volatility and currency noise. Trade implication: keep gross down, keep convex hedges on, don’t fund longs with fragile leverage. The sense of urgency came after leaders at Munich spoke in “great‑power politics” terms, with escalation paths that start in trade/tech/capital wars and can end in military war. Translation for risk: raise the geopolitical volatility premium, shorten time horizons, and stop pretending capital controls/sanctions are “EM only.”
Even if nothing “breaks,” the market has to price the tail: Iran is already signaling leverage at sea, including a temporary Strait of Hormuz closure for live‑fire drills (the chokepoint that moves ~20% of global oil). On top of that, the U.S. Supreme Court struck down Trump’s sweeping “reciprocal” tariffs. For context, tariffs are taxes on imports. In the U.S., Congress owns taxing/tariff power. A president can only do it if Congress clearly gave that power away in a specific law. The Court also leaned on the idea that if a decision is massive economically/politically, Congress has to say ‘yes, you can do this’ in plain terms.
Crypto: The Market Isn’t Ready Yet
Price is behaving like a market that’s still de‑risking: BTC is repeatedly failing around $70k, with flows looking less like “fresh bid” and more like tactical positioning. The biggest short-term flow datapoint remains ETF behavior, but recent sessions have shown meaningful outflows (e.g., one day last week with reported ~$410M net outflows across U.S. spot BTC ETFs), which matters because ETFs are the cleanest “institutional marginal buyer” channel.

Across crypto, leverage is already materially cleaned up (BTC OI down massively vs Oct 2025 peak; current BTC OI ~$44.8B), so squeezes need fresh flow, not just “positioning is light.” More specifically, BTC is the “fiat liquidity fire alarm,” and its divergence vs a flatter Nasdaq is a warning sign of what may be coming next: USD-credit deflation > AI-induced white-collar displacement > consumer/mortgage defaults > Fed money printing.

While divergences can be just a data point within a slower trend, the directional setup is still there: BTC vs QQQ (divergence widening = stress) + gold up vs BTC down = confirmation of “deflationary bid” behavior, which can be reinforced by software/SaaS underperformance.
AI Digest: The Compute Arms Race
Familiarize yourself with this term: Cournot Equilibrium. This is the economic phenomenon that describes the current AI lab “arms race” as economically rational behavior: a few frontier labs overspend on training to secure scarce, high‑margin inference capacity, and prices emerge from that, not from cost‑plus logic or textbook valuation metrics like P/E.
There are two intertwined businesses hiding in each lab’s P&L: (1) the lumpy, high‑risk training pipeline (R&D capex), and (2) the smoother inference “factory”; any valuation or strategy work that doesn’t separate these will misread profitability and risk. Training is treated as a massive upfront capex that creates a model “asset” which rapidly depreciates as the frontier advances, while inference is essentially a manufacturing business with healthy contribution margins driven by subscriptions, API usage, and enterprise deals.
This explains why Labs appear insanely unprofitable today, but this is a rational bet on capturing future high‑value demand from knowledge workers and enterprises who are willing to pay heavily for the very best models. As an example, training GPT‑3.5 cost on the order of $100 million, which at the time felt expensive. But once GPT‑4 launched and ChatGPT subscriptions and API usage exploded, the inference business quickly reached a multi‑billion‑dollar annual run‑rate, so the original training cost was earned back and then some.
The key strategic question becomes: does this converge to a stable oligopoly with durable high margins on frontier access, or does expanding supply and substitute models drive a future price war that crushes returns?
- Dario Amodei is still selling “fast takeoff” (“geniuses in a data center” within a few years, potentially sooner) while simultaneously conceding diffusion and verification bottlenecks.
- Peter Steinberger, creator of OpenClaw (former ClawdBot), joins OpenAI. The project will live on as an open source project that OpenAI will continue to support.
- Anthropic ships Sonnet 4.6, its most capable Sonnet model yet. With a 1M context window, it seeks to achieve Opus-level performance across coding, computer use, long-context reasoning, agent planning, knowledge work, and design.
- Grok 4.20 came out with a multi-agent architecture (not MoE or serial processing) where four specialized AI agents debate in real-time.
- AI‑led disruption fears remain an overhang for mega‑cap tech multiples.
- Bytedance disrupts Hollywood while moving to curb unauthorized IP use on its AI video tool after a Disney threat (expect more “guardrails” headlines and product throttles).
- The UK is signaling a crackdown on AI chatbots tied to child safety.
- EVMBench by OpenAI and Paradigm seeks to measure how capable AI agents are at smart contract security across three modes: Detect (find known vulns), Patch (fix without breaking intended behavior), and Exploit (execute end‑to‑end fund‑draining attacks) in a sandboxed EVM environment.
Analysts still flag the “$1T revenue by decade‑end” talk as delusional, arguing that price competition kills moat, and warning the capex boom can overbuild like 2006 housing / dot‑com fiber even if the tech is real. On the opposite end of the “AI bubble” spectrum is the “everything is compute” thesis, whose pitch boils down to: intelligence scales fastest where energy is transformed into compute which is then transformed into automation that compounds, and legacy institutions can’t compress reality fast enough. If you believe in the narrative, trade it as a regime statement, not a one-week catalyst: own infrastructure rails on pullbacks; fade beta when liquidity tightens.
Signals Radar: Stay Light, Stay Liquid
BTC — Price is back in the high‑$60ks, but still failing to sustain >$70k on closes. Spot ETF flows have been choppy-to-negative with big red prints (e.g., Feb 11 -$276.3M, Feb 12 -$410.2M, Feb 17 -$104.9M) and tiny green days (Feb 13 +$15.1M). Treat $70k as the “prove it” level; only like adding risk if we get consecutive positive ETF days and reclaim/hold >$70k post Friday’s PCE print. Otherwise, fade strength into $70k and buy dips only at prior demand ($66–67k) with a hard line under ~$65.8k (if that breaks, the market is telling you the bid is gone, not early).

ETH — Flirting around $2k, but still trading like “BTC beta minus 1” (needs an actual catalyst to lead). ETF flows are improving at the margin after heavy outflows of >100M. As it stands, ETH is an asset that only earns size if it can hold >$2,050–2,100 while ETF flows remain net positive; otherwise it’s a range trade with $2,000 as the obvious “don’t fight it” line.

SOL — SOL/BTC is the key variable to track. SOL is a “buy dips” as long as it holds $80–83; if it loses that zone and onchain volumes keep dropping, you cut (SOL outperformance without activity usually doesn’t last).

HYPE — Basically flat on the week, but the internals matter more than the spot chart as the project is now proactively trying to remove offshore/illegal overhang risk in the U.S. Perps volume is still huge and OI remains elevated. It’s definitely one of the few chains and protocols where liquidity is sticky even as volume cools down. Importantly, the Hyperliquid Policy Center was introduced in Washington, D.C., with Jake Chervinsky as CEO, explicitly aiming to help build a legal framework for DeFi and perps in the U.S. A Hyperliquid‑tied foundation is donating 1,000,000 HYPE (~$28M), and the founding team includes Brad Bourque and Salah Ghazzal; they’re already hiring core policy roles. In other news, Tether invested in Supreme Liquid Labs (Dreamcash parent) to push USDT0‑collateralized perpetuals on Hyperliquid via Dreamcash’s frontend, including markets like TSLA, NVDA, S&P 500, gold, and silver. Size undisclosed.

DOGE — Volatility kicked off again with a Musk headline reigniting the old “literal DOGE on the moon” meme (replying “maybe next year”), as SpaceX shifts priorities to building a self-growing city on the Moon, which they think they can potentially achieve that in less than 10 years, whereas Mars would take 20+ years.

ZRO — After last week’s Zero announcement, Google Cloud joins forces as a strategic partner to explore how AI agents could make micropayments and trade resources instantly without a bank account.

AAVE — Aave is in a real governance standoff over who controls the Aave brand + front-end distribution and who captures “offchain/product” revenue. After tokenholders challenged Aave Labs (Stani’s side) for monetizing swap/interface flows and effectively diverting revenue that previously accrued to the DAO, the debate escalated into broader demands that the DAO secure ownership/stewardship of key brand assets via a legal wrapper/Foundation. On Monday, Aave Labs proposed a “reset” framework where 100% of Aave-branded product revenue would flow to the DAO (including swap revenue they estimate at ~$10M annualized) but in exchange the DAO would fund Labs with a large annual budget and grant AAVE tokens, while Labs retains execution autonomy. Marc Zeller/ACI says the direction is right (revenue alignment + Foundation + V4), but is pushing to unbundle the omnibus vote, define/audit “100% revenue” so it’s enforceable, ensure the Foundation is truly independent, and require governance wallet disclosures before transferring more voting power, while Stani opposes additional “soft governance” norms and argues token voting sovereignty should remain the core mechanism. Finally on Friday, BGD Labs (one of Aave’s core, long‑running independent DAO service providers doing protocol maintenance, governance tooling, and security coordination) announced they will not renew and will cease contributing when their current engagement ends on April 1, 2026. Their stated reason is the ongoing alignment fight with Aave Labs, specifically what they describe as an organizationally asymmetric setup (brand/comms control + voting power) and an adversarial push to deprecate v3 in favor of v4 without real collaboration.

PUMP — Releasing “cashback coins” (users choose creator fees), letting users decide whether a token keeps creator fees or flips rewards toward traders (“Cashback Coins” toggle in‑app). Basically, incentive engineering: attract activity by paying the churn. In the trenches, expect faster rotations and more mercenary volume.

JUP — Introduced its net-zero emissions initiative, with a vote ending on Feb 21st. This would pause emissions from the Team Reserve indefinitely, absorbing any team sales of vesting tokens in the Jupiter treasury (if the team sells, it would come from the treasury, fully visible onchain), and postponing Jupuary indefinitely. One of Jupiter's founders also hinted at more upcoming JUP utility within products.

ONDO — Despite the large FDV, this is one of the cleanest “RWA + distribution” plays right now. Ondo launched Ondo Perps (equity perpetual futures, up to 20x leverage, non‑US) and Ondo Global Listing (tokenize NYSE/Nasdaq IPOs day‑one). MetaMask integrated Ondo Global Markets so eligible non‑US users can trade 200+ tokenized US stocks/ETFs/commodities inside the wallet. Ondo also cites EU regulatory approval covering 30 European markets / 500M investors (passporting style expansion).

OP — Just weeks after the activation of the Superchain’s buybacks program using OP Stack sequencing fees, of which Base represented the large majority, Coinbase went on to abandon the OP Stack to build their own, consisting of their own spec, code, and infra.

LIT — Since announcing that to access the Lighter Liquidity Pool (LLP) you must stake (Jan 14 launch + Jan 28 enforcement), with 1 LIT staked unlocking up to 10 USDC of LLP quota and a 3‑day unstake lock, staking progress is now at ~46% of the airdrop (near 50%). Only ~25% float unlocked (~250M LIT), with 46% of that (115M LIT) effectively pulled out of liquid circulation via staking means that float compression can matter if demand stays constant.

WLFI — After growing its USD1 stablecoin supply with “World Liberty Markets” launched 4 weeks ago, they are now looking for more distribution with a new forex/remittance platform pitched at lower fees, “World Swap.” This is politicized payments. It can drive attention/flows, but it also imports headline risk and conflict‑of‑interest scrutiny into crypto rails. Treat it as a volatility catalyst, not a “fundamental” bid you can underwrite.

DOLO — While fundamentals have stepped up hard: TVL ~$202M (~5.5x TVL/FDV), ~$12M annualized fees (~3.1x P/F on FDV), and borrowed value ~9x since Jan 12, 2026 when WLFI’s white‑label lending stack (“World Liberty Markets”, USD1‑first + points) shipped on Dolomite, the market is treating it as “sell‑the‑news + mercenary incentives” despite the business inflecting. The real driver is USD1: supply ~$5.3B (+73% MoM), and borrow caps already binding (~$100M borrowed / ~$160M supplied vs $1B cap) with users loop‑farming ~4.88% WLFI incentives on top of ~2.6% supply rate + points; that same incentive stack is the main flaw: if incentives roll off and TVL/borrows mean‑revert, the “cheap multiple” won’t hold. More on our latest memo, where we note how its utilization is sector‑leading at 67% (vs Aave 64% / Euler 59% / Morpho 36%), but risk is concentration (stablecoins are 90% of borrows; USD1 ~40% of borrowed value) and competition (USD1 onboarding on Aave has passed a temperature check). Value accrual is the unlock: docs point to a DAO‑driven fee switch + borrow discounts via veDOLO (plus oDOLO mechanics); but today revenue doesn’t flow to holders and the team reported being unprofitable in Q4‑25, with team unlocks starting May (team owns 20.21% of supply) = real supply overhang risk.

APT — Tokenomics upgrade: from “bootstrap inflation” to performance‑driven supply. they plan to push governance changes that cut staking rewards from ~5.19% to 2.6%, raise gas fees 10x (all gas fees paid in APT are burned; they claim stablecoin transfers still land around $0.00014), and introduce a hard cap of 2.1B APT plus a permanent Foundation lock/stake of 210M APT (framed as burn‑equivalent supply removal vs sale/distribution). The core bet is throughput: Decibel, a fully onchain orderbook DEX, is positioned as the burn engine. Aptos projects >32M APT/year burned as Decibel scales (100+ markets), alongside performance‑gated grants and an “explore” path for a programmatic buyback reserve.

DBR — "Agents need onchain rails” catalyst (MCP) with tradable momentum risk. When risk-on is prevalent, the “agent narrative” generally pumps fast, but then mean‑reverts harder. How to play: only pay up if you see follow‑through integrations (wallets/agent frameworks shipping DBR routes). Invalidation: it stays a press cycle with no distribution (no real usage).

RPL — The Saturn I upgrade is the main event, introducing changes in tokenomics and bringing more capital efficiency. Operator bond halves to 4 ETH, MEGAPOOLS consolidate many validators under one contract to reduce gas/ops friction, and the protocol-wide “fee switch” redirects a portion of protocol ETH staking rewards toward staked RPL (i.e., pushing RPL away from “mostly inflation incentives” toward ETH-revenue accrual), with the DAO able to adjust revenue splits dynamically.

EIGEN — Well-positioned for a world where an increasing number of people buy into the idea that crypto was built for AI agents and not humans clicking buttons and signing approvals. Crypto UX is hard, and EigenCloud’s rebranding was their first step to reclaim the lead as “the verifiable cloud for the agentic era”: as AI agents start touching real value (money, reputation, private data, IP), the core requirement shifts from “fast/cheap compute” to provenance + proof + accountability. Concretely, they now see themselves as “infrastructure-complete” across four primitives: EigenLayer (now “protocol-complete with slashing + redistribution”), EigenDA (testnet throughput milestones for internet-scale data availability), EigenCompute (AWS-like developer experience but with onchain verifiability + smart contract bindings), and EigenAI (deterministic / verifiable inference for open-source models). This has given them a head-start on integrations and partnerships with giants like Google and Coinbase, and argue this is not a pivot but the same pattern as AWS shipping primitives (S3/EC2/etc.) that later unlocked an application universe.

BIRB — Korea listing flows + Adam Weitsman publicly adding 200 Moonbirds NFTs (mindshare fuel) remind us that fast money is still a thing, and that there are opportunities to be captured even in the absence of long-term conviction. These are opportune setups for momentum longs and fading breakdowns.

Prediction Markets — Volumes keep making new highs across venues, proving their ability to become the “everything casino” while crypto prices are plummeting. Vitalik’s critique: prediction markets over‑optimize for short‑term gambling + “finding dumb money,” which degrades the product.

Conclusion
This is still a de‑risking regime: policy uncertainty up, geopolitics loud, and crypto missing a clean marginal buyer. Move too early and you burn capital building exposure before the market is ready; move too late and you’re defending yesterday’s margins. Disinflation is improving, labor isn’t cracking, but the Fed is telling you policy volatility is still on the table. So we stay light, keep convexity, and force the market to confirm before we size.


