Warsh's first FOMC: the Fed chair who refused to draw his own dot — and launched five task forces to redesign the Fed itself.
The 12-0 hold at 3.50–3.75% was unanimous. The surprise was everywhere else: the median 2026 funds rate projection jumped 40bp to 3.8% (implying at least one hike), the policy statement was cut from 341 to 130 words with easing-bias language deleted, and Warsh — the new chair — declined to submit his own dot plot on principle. He then announced five task forces to overhaul Fed communications, balance sheet, data, productivity/jobs, and inflation. CME FedWatch flipped: bets on a year-end hike surged from 15.7% pre-meeting to >40% post. The easing trade is dead.
+40bp
Dot plot shift · 3.4% → 3.8%
40%+
CME hike odds · from 15.7%
−62%
Policy statement length (341→130)
5
Task forces to overhaul Fed
SPCX$191.82
SPCX Δ-4.95%
S&P 500-1.21%
Nasdaq-1.34%
10Y Yield+6bp · 4.49%
Gold-1.77% · $4,254
Fed Hold12-0
Hike Odds>40%
CuspAI$2.6B
GLM-5.2744B MoE
The Hawkish Surprise
The 2026 dot plot jumped 40bp — and 9 of 18 saw a hike
March's median projection was 3.4% — pricing in two cuts by year-end. Yesterday's median: 3.8%, with 9 participants penciling in at least one hike, 1 a cut, and 8 no change. Median 2026 PCE inflation projection: 3.6% (up from 2.7% in March). The shift is large enough to invalidate the entire H2 2026 easing trade.
Mar 2026 · median
3.4%
Jun 2026 · median
3.8%
SEP PCE inflation
3.6%
Core PCE
3.3%
Unemployment
4.3%
GDP 2026
+2.2%
All projections from June 17 SEP. Warsh declined to submit his own dot — 18 of 19 participants submitted.
Why Warsh Refused to Draw
"Not helpful in the conduct of policy"
The new Fed chair, confirmed 54-45 and sworn in May 22, skipped his own dot on principle — and the statement word-count tells you what he thinks about committee consensus language.
Cut statement 341 → 130 words
Deleted easing-bias language entirely
Launched 5 task forces
Declined personal dot
First hike now seen Oct
Sustained 12-0 vote
Five Task Forces to Overhaul the Fed
Warsh didn't just change rates — he started redesigning the institution
Each task force is a separate reform track, with new staff and external consultants. If the communications task force scraps the dot plot within 12 months, every rate-vol model has to be rebuilt. The vol-of-vol trade re-opens.
Task Force 1
Communications
Statement, dot plot, presser format. Could scrap the SEP entirely.
Task Force 2
Balance Sheet
QT pace, MBS runoff, reserve adequacy. Could end run-off in H1 2027.
Task Force 3
Data Sources
Replace/supplement BLS with private + alternative data feeds.
Task Force 4
Productivity & Jobs
Re-examine labor market signal quality post-AI displacement.
Task Force 5
Inflation
Re-anchor 2% target or replace with average-inflation targeting.
The structural read: Warsh is using his honeymoon period — before the first real policy mistake — to rebuild the plumbing. By the time cuts or hikes matter again, the Fed will look like a different institution.
Markets Already Priced the Pivot
Five moves in 24 hours, all aligned
The tape validated the hawkish shift instantly. Every asset class moved in the direction Warsh implied.
S&P 500
-1.21%
Nasdaq
-1.34%
Dow Jones
-0.98%
10Y Yield
+6bp · 4.49%
Gold
-1.77%
All intraday Jun 17 closes. CME FedWatch: 15.7% → 40%+ on a 2026 hike.
The File-as-Manifest Convergence
Vercel Eve + Cloudflare Flue shipped the same day. Same six primitives. The agent framework is now a commodity.
Both launched June 17 — within hours of each other. Both treat an agent as a directory of files (Eve: filesystem-first TypeScript with agent.ts config; Flue: Astro-team-built on Cloudflare's Agents SDK, one Durable Object per agent). Both ship identical primitives. Cloudflare's framing: "2026 is the year agent harnesses go to production." Vercel's: "production already built in."
Why it matters: The agent is becoming a commodity primitive. Value moves to the platform — the cloud, the data plane, the integration, the procurement relationship. Same pattern as Lambda functions in 2015. Whoever owns the file-as-manifest runtime owns the abstraction layer for the next 5 years.
Vercel Eve
Filesystem-first
Cloudflare Flue
Durable Objects
Hermes Agent (TTL)
In this paradigm
Hermes Agent already operates file-as-manifest. TTL's R-405 (control-plane land grab) is the layer underneath — the boring infrastructure these harnesses sit on.
Six Primitives · Eve & Flue
Same list, two vendors
1 · Durable Exec
Workflow / DO
Resumable across crashes
2 · Sandbox
Eve Sandbox / MicroVM
Isolated code exec
3 · MCP + OAuth
Tool auth standard
Every framework
4 · Multi-channel
Slack/Discord/Twilio
8+ channels
5 · Approvals
HITL gates
First-class
6 · Evals
Regression suite
CI/CD-grade
Open-Weight Coding Frontier
Z.ai GLM-5.2: 744B MoE, 1M context, trained on Huawei Ascend
Released Jun 13, fully detailed Jun 18. 40B active params (MoE), 128K output tokens, 1M-token context. Scores 81.0 on Terminal-Bench 2.1 (vs 62.0 for GLM-5.1) and 62.1 on SWE-bench Pro. Trails Claude Opus 4.8 by 1pp on FrontierSWE. IndexShare sparse attention (2.9x FLOPs reduction at 1M context). MIT license. Day-one API on Claude Code, Cline, OpenCode, Roo Code, Goose, Crush, OpenClaw, Kilo Code.
744B total / 40B active
1M context · 128K output
Trained on Huawei Ascend 910B
MindSpore framework, no NVIDIA
+19pp jump on Terminal-Bench
MIT license · open weights
SPCX First Post-IPO Drop
$191.82 (−4.95%) — Options opened, bears got their first lever
The first full-session decline since IPO. Driver: SPCX options began trading Tue Jun 16. Day-1 volume ~1M call contracts. Puts now available. The 3-day +47% post-IPO run was technically overextended and structurally protected by float mechanics — options opening is the natural air-pocket.
80%
Lockup test probability
Extra 10% of insider shares only unlocks if SPCX closes ≥$175.50 on 5 of 10 sessions before Q2 earnings. Currently $191.82 — well above threshold.
The World-Model Funding Wave · 2026 YTD
Bezos's third physical-AI move in two weeks — the category that didn't exist as a $1B+ class 12 months ago
🧠 Prometheus
$12B · $41B post
⚙️ CuspAI
$400M · $2.6B post
🌍 Odyssey
$310M · $1.45B post
🎬 Runway
$315M · $5.3B post
🎮 Decart
$300M · $4B post
🤖 AMI Labs (LeCun)
$1.03B
🧪 World Labs (Fei-Fei)
$230M
Total raised 2026 YTD
>$2.2B
Bezos's three moves in two weeks: Prometheus $12B (Jun 12, design-to-manufacturing engineering AI), CuspAI $400M (Jun 17, materials design AI), plus his share of Odyssey via AWS exclusive + Trainium (Jun 17). Bezos is now the anchor LP of physical-world AI. CuspAI's valuation went from $30M seed → $100M Series A → $2.6B in 9 months. Materials design specifically (CuspAI, Orbital Materials, Periodic Labs) is becoming its own $10B+ sub-vertical.
TTL Strategic Read
Three reads from today's tape
1 · The easing trade is dead
Warsh's dot plot kills the H2 2026 cut narrative. For TTL: a sustained hawkish Fed pulls rate-cut tailwind out of the 2027 IPO window — Anthropic and OpenAI public-offering timing gets harder, not easier. Recalibrate client projections that assumed a Sep cut.
2 · Eve + Flue = agent = commodity
When Vercel and Cloudflare ship identical six-primitive frameworks on the same day, the agent layer is a commodity. TTL's bet on Hermes Agent + control-plane (R-405) is correct — the platform layer underneath is where the next 5 years of value accrues.
3 · Open-weight catches up fast
GLM-5.2 at 81.0 on Terminal-Bench (vs Opus 4.8 by 1pp) at a fraction of commercial cost = the procurement math just flipped. For TTL consulting: smaller model + skills = smaller model alone + closed frontier. Re-anchor any client-side model-selection rationale this week.
This Week · June 18–24
Binary events ahead
Jun 17 · Tue
Warsh FOMC · Hawkish dot, 5 task forces. Done.
Jun 18 · Wed
Anthropic Glasswing rebuttal vs export-control order