ANDREW: Good morning. Saturday, April eleventh, twenty twenty-six. You're listening to Morning Signal. I'm Andrew.

AVA: And I'm Ava. Andrew, I have to say — I read through the full briefing last night and I did not sleep well.

ANDREW: That's... actually kind of the appropriate response.

AVA: I mean it. There's a lot going on.

ANDREW: There is a lot going on. And what's unusual about this particular moment — and I want to dig into this — is that all of it is connected. Like, the stuff that's happening in the Middle East, the AI buildout, the Fed being stuck... these aren't separate stories right now.

AVA: They're feeding each other.

ANDREW: They're feeding each other. Exactly. So let's start with markets and macro because I think that's the scaffolding you need to understand everything else. Ava, you want to set the scene?

AVA: Sure. So the baseline here — and this is coming out of the Forward Guidance pod this week — Hormuz has been functionally closed for forty-five plus days. Oil is sitting at a hundred dollars front-month. The ten-year Treasury is at four point two seven percent. And core PCE has been running at forty basis points a month for the last three months.

ANDREW: Which annualizes to almost five percent.

AVA: Right. And here's the thing that Forward Guidance flagged that I keep coming back to — they're saying there are three or four more "scalding" prints still ahead of us in this data cycle. So whatever pain we've felt on inflation so far... we may not be done.

ANDREW: And the Fed can't really respond to any of it because — and this is from RenMac Off-Script this week, they had Rich Clarida on, former Fed Vice Chair — Kevin Worsh can't get confirmed because Senator Tillis is blocking the vote until the DOJ drops its criminal investigation of Jay Powell.

AVA: So you have this policy vacuum right at the exact wrong moment.

ANDREW: The exact wrong moment. And Clarida was actually pretty specific about the timeline. His modal case is that Worsh chairs the June FOMC meeting. But there's this wild wrinkle — and this came out of RenMac too — the FOMC chair is elected by the committee. So Powell could theoretically stay on as governor and just... be in the room. Like this shadow Fed chair dynamic.

AVA: That would be genuinely bizarre.

ANDREW: Eccles and Truman era precedent apparently. But bizarre, yes. Now, look — on equities. The Compound and Friends this week had Steve Sosanik from Interactive Brokers on, and the framing I thought was really clarifying was this: the S&P is only about three percent off all-time highs. Max drawdown since February twenty-seventh is about nine percent. And yet forward PE has compressed eighteen to nineteen percent over that same period.

AVA: So earnings estimates haven't moved.

ANDREW: Haven't moved. JP Morgan's Guide to the Markets shows consensus twenty twenty-six EPS at or near all-time highs. So what you have is a bear market in multiples that's being masked by simultaneously rising earnings estimates. And those two things cannot both be right for long.

AVA: And Sosanik had that great line — something like, S&P three percent off all-time highs with Hormuz closed forty-five days, oil at a hundred, core PCE reaccelerating... "you would take that short blindfolded every day of the week."

ANDREW: That's the bear case. The bull case — also from Forward Guidance — is basically extreme positioning unwind. Goldman's derivatives desk is saying the systematic community is still net short thirty-seven billion dollars of US equities. CTAs are projected to buy forty-five billion over the next week in a flat tape. And the contrarian framing is: Trump needs midterm wins, and midterms are eight months out. So there's a policy put under markets.

AVA: So bearish that you're bullish.

ANDREW: So bearish that you're bullish. Which is either very sophisticated or deeply insane.

AVA: Probably both. What was the retail data like?

ANDREW: So this was interesting from The Compound and Friends — Citadel data showed the first week of net retail selling since November twenty twenty-five. And it happened to coincide exactly with the day the Fed expectations flipped from pricing a hike to pricing a December twenty twenty-six cut. Which tells you retail was watching the Fed, not the geopolitics.

AVA: And what were they actually buying on the dip?

ANDREW: This is the most behaviorally interesting part. The most-bought instrument at Interactive Brokers during the drawdown was VO — the Vanguard 500 ETF. Not leveraged products. Micron and Nvidia were bought on dips. Tesla bought on selloffs. Microsoft... steadfastly bought despite essentially no bid from the market.

AVA: So people are buying the companies they believe in, not trying to trade around the chaos.

ANDREW: Basically. Okay, let me flag one more macro thing before we move to tech, and this is from RenMac Off-Script — Clarida was explicit that AI capex right now is not a counterbalance to inflationary pressure. It's adding to it. His framing was that near-term, all this buildout is demand-side stimulus in "old economy directions." Construction workers, concrete, power plants, Nvidia chips. The supply-side productivity benefits from AI are... "not there yet."

AVA: Which is such an important point because I think the narrative in markets has been, AI is this deflationary force that the Fed should be leaning into. And Clarida is basically saying no — not yet. Maybe eventually, but right now it's inflationary.

ANDREW: Right now it's inflationary. And that changes the rate math considerably.

AVA: Okay, can I pivot to tech? Because the Anthropic story this week is where all of this gets really interesting and also really contested.

ANDREW: Please.

AVA: So over on the All-In Pod this week, they spent a huge chunk of time on the Anthropic "Mythos" model release. Released the week of April ninth. Apparently it's blown up expert benchmarks. But here's the key thing — Anthropic declined to release it publicly.

ANDREW: Which immediately raises the question of why.

AVA: Right. And what they're claiming — again from All-In — is that this model can autonomously identify thousands of OS and browser vulnerabilities. And not just find known ones. They're talking about a twenty-seven-year-old OpenBSD vulnerability used in critical infrastructure firewalls. A sixteen-year-old FFmpeg bug that survived five million automated scans. Novel exploit chains that combine three to five independent vulnerabilities.

ANDREW: Wait, how does a bug survive five million automated scans?

AVA: That's exactly the question. Because automated scans are looking for known patterns. Something genuinely novel doesn't fit the pattern. Which is the claim here — that Mythos is actually doing creative security research, not just pattern matching.

ANDREW: And they've assembled... what is it, forty companies?

AVA: So this is Project Glass Wing. Forty companies — Apple, Microsoft, Google, Amazon, JP Morgan named — in a hundred-day sandboxing period. Find and patch vulnerabilities before public release. Completely voluntary, no government mandate.

ANDREW: Which Sacks on All-In acknowledged is "more on the legitimate side" — his words — even though he also pointed out that Anthropic has a documented pattern of fear-based marketing. He mentioned that blackmail study where they ran two hundred plus prompted runs to get the result they wanted.

AVA: Chamath was more skeptical. His read — and this is still from All-In — was basically, remember GPT-2 in twenty-nineteen? One point five billion parameter model, hyped as a civilization-ending risk, turned out to be nothing. He's essentially applying that frame here. And he said something that stuck with me: "an advanced hacker can do this today with Opus."

ANDREW: Which is either true or terrifying depending on which part of "an advanced hacker" you focus on.

AVA: Ha. Fair. But the market clearly didn't wait for the debate to resolve. Palantir down eight percent, Palo Alto down five, CrowdStrike down eight, ServiceNow down eight — consecutive sessions.

ANDREW: Which is interesting because the Forward Guidance read on this was actually different from the security angle. Their framing was: this is really a compute supply story disguised as a policy decision. If Mythos is a national security asset, AI compute investment becomes militarily non-discretionary. Which creates a pathway for government funding regardless of macro conditions.

AVA: The hardware thesis in other words.

ANDREW: Exactly. And that maps to the semis-versus-software divergence that The Compound and Friends has been tracking — SMH within one percent of all-time highs while IGV is breaking below daily, weekly, and monthly moving averages simultaneously. Dell, HP, NetApp, Super Micro going vertical. Software getting crushed.

AVA: And Sosanik pulled out what I thought was the single most arresting data point — on two consecutive days, the S&P rose three point one percent while IGV fell five point two percent. That combination had never occurred in the prior twenty-five year dataset.

ANDREW: Never. Once.

AVA: Which either means we're in genuinely unprecedented territory or something is very wrong with how either semis or software is being priced.

ANDREW: Both possibly. Okay, what about the OpenAI versus Anthropic competitive picture? Because Big Technology this week had a lot on this.

AVA: So Big Technology this week — the revenue numbers are remarkable. OpenAI is at roughly twenty-five billion annualized. Anthropic is at about nineteen billion. And the trajectory implies a revenue crossing point within twelve months.

ANDREW: Which would have seemed insane eighteen months ago.

AVA: Wild. And OpenAI's response to Claude Code's enterprise momentum — again this is Big Technology — is apparently to kill what they're calling "side quests." Sora is on the chopping block. The pivot is to a super app bundling ChatGPT plus Codex plus this Atlas browser they're building. Basically trying to own the surface area.

ANDREW: The OpenClaw thing from All-In was also pretty telling. Anthropic terminated two-hundred-dollar-a-month subscriptions from users who were consuming two to twenty thousand dollars in tokens. Which — fair enough from a unit economics standpoint. But then within ten days they announced Claude Managed Agents on April eighth. And Sacks read that as: "They systematically copied feature by feature, incorporated it into Claude, then cut off the oxygen."

AVA: And Chamath took the antitrust framing further. If Anthropic provides first-party agents at flat rate while third parties pay metered, and Anthropic has fifty to sixty percent of coding token market share — his estimate — that's potentially bundling under antitrust law.

ANDREW: Interesting. Okay here's the thing that gives me pause on all of these valuations though. The BitTensor story from All-In.

AVA: Oh, this one.

ANDREW: BitTensor Subnet 62 — they're calling it Ridges AI — spent approximately one million dollars in TAO token rewards. Hit eighty percent of Claude 4 capability in forty-five days. Via anonymous distributed contribution, Bitcoin-mining-style incentives.

AVA: One million dollars.

ANDREW: One million dollars versus what OpenAI reportedly spent getting to their current capability. And if open source can close capability gaps that quickly...

AVA: Then the terminal value case for an eight-hundred-and-thirty-billion-dollar private valuation depends entirely on that remaining twenty percent gap being the part that matters.

ANDREW: And Chamath's point was that the bottleneck isn't frontier capability right now. His number was five percent enterprise AI adoption at the broad market level. One company he mentioned still brings in sixty-year-old retirees to interpret COBOL. So the frontier gap may not be what's constraining deployment.

AVA: Which makes the IPO math complicated. The Compound and Friends flagged that SpaceX, OpenAI, and Anthropic are all potentially targeting similar market windows. And for S&P index inclusion alone, Invesco and peers would have to sell ninety-nine other positions to fund purchases. The supply-demand equation that's been a tailwind for equities for a decade actually reverses.

ANDREW: Forward Guidance host predicted neither OpenAI nor Anthropic would actually IPO in twenty twenty-six. And apparently OpenAI's CFO Sarah Frier was excluded from investor meetings on server spending — which is... a tell.

AVA: That's a tell, yes. Okay, let's get to geopolitics because I think we've been circling it and it's really the foundation of everything.

ANDREW: Yeah. Let's do it. What's the on-the-ground picture?

AVA: So Geopolitical Cousins this week had the most granular operational breakdown I've seen. Israel has struck Beirut — two-hundred-fifty-plus killed, a thousand-plus injured. Drones hit Saudi pipelines. Reports of Abqaiq on fire. US conducting twenty percent of daily strikes on Iran, Israel eighty percent. And there is technically a ceasefire in place, but...

ANDREW: But.

AVA: Jacob on Geopolitical Cousins described it as "apparently in ceasefires you still throw bombs at each other." Iran's parliament speaker has cited three violations already. Lebanon is the real flashpoint — Jacob sees it as a genuine Iranian red line, while Marco thinks it's something Iran will ultimately sell for economic reasons.

ANDREW: And the tolling mechanism — this is what I keep coming back to from Geopolitical Cousins — because I think Marco's framing here is the most intellectually interesting and most underpriced thing in markets right now.

AVA: Explain the tolling thing for people who haven't heard it.

ANDREW: So the IRGC has apparently set up a formal system for ships to pay to transit. And I mean formal — reportedly a Delaware LLC, an online portal, two million dollars per vessel, accepting Bitcoin, yuan, and riyals. And Marco's point is that incremental vessel traffic was already ticking up before any ceasefire, because some ships are just paying.

AVA: So the binary framing that markets are using — Hormuz open versus Hormuz closed — is wrong.

ANDREW: May be exactly wrong. Because if Marco is right, the ceasefire collapsing doesn't necessarily produce the oil spike the market fears. And the ceasefire continuing doesn't necessarily reopen Hormuz fully. The actual variable to track is the tolling mechanism's stability, not the ceasefire headline.

AVA: And The Compound and Friends' black-swan model — which had oil going to one-fifty to two hundred on a full closure — would be wrong not just in magnitude but in its entire framing.

ANDREW: Exactly. Now the Hormuz arithmetic from Geopolitical Cousins is worth knowing. Normal traffic is eight hundred to a thousand ships per day. Current actual throughput is thirty-five to forty. And only about two VLCCs — four million barrels — need to transit to cover the near-term supply gap. But that number rises to four VLCCs minimum by end of April. So the window is tightening.

AVA: And the commodity dependencies go way beyond oil. This was in both Geopolitical Cousins and The Compound and Friends — helium sourced from the Persian Gulf is a critical input for semiconductor fabrication. Fertilizer downstream is affecting food supply for three billion people in Pakistan, India, and China. These are not priced.

ANDREW: They're absolutely not priced. And the Iran-China relationship from Geopolitical Cousins is one of those things that looks simple and isn't. Iran runs a trade deficit with China despite being China's third-largest crude supplier. China is Iran's economic lifeline including drone components. So China has leverage over Iranian behavior. But simultaneously, prolonged Hormuz disruption threatens the LNG going to South Korea and Taiwan that powers the semiconductor fabs feeding China's AI supply chain.

AVA: Right. So China is this inadvertent winner in fixed income — Forward Guidance noted that thirty-year Chinese bond yields were the only major yields that didn't rise post-Iran war — but also an inadvertent victim through the supply chain.

ANDREW: And they're implicated in the conflict through Iranian drone components. Geopolitical Cousins confirmed Iranian drone supply chains contain Chinese, Korean, and reportedly US-origin components. Which means sanctions enforcement is confirmed porous, and China has secondary sanctions exposure if this escalates.

AVA: The Forward Guidance read was "China may be winning." The Geopolitical Cousins supply chain analysis is... more complicated than that.

ANDREW: Substantially more complicated. Okay, let's close the loop with private credit because I think it's the risk nobody's really pricing.

AVA: Yeah. Invest Like the Best this week had Alan Waxman from Sixth Street on and it was one of those conversations where someone really careful says something alarming very quietly.

ANDREW: What was the alarming part?

AVA: So the setup is this — private credit has grown from roughly five hundred billion pre-GFC to about two trillion today. FRE multiples have gone from ten to fifteen times in the early two thousands to twenty-five to thirty-plus currently. And some of these perpetual private BDCs are seeing redemption requests exceeding the five percent quarterly limit. So you have illiquid assets paired with semi-liquid terms.

ANDREW: And Waxman's line was —

AVA: "Semi-liquid is not a real thing." That was the line. His diagnosis is that the root cause is the "factory model" of private capital — pushing retail wealth channel money into principal risk-taking next to illiquid assets. Same asset-liability mismatch as the GFC, just in unregulated clothing.

ANDREW: And The Compound and Friends made explicit what makes this directly connected to tech — Sosanik said "a lot of the software boom was financed with private credit." So if software EPS rolls over in Q2 earnings — which IGV's chart suggests is coming — equity and credit holders get hit simultaneously into a private credit market that's already under redemption pressure.

AVA: Waxman said "not yet systemic" — but that qualifier depends on the economic backdrop remaining decent. And decent is a hard thing to guarantee with oil at a hundred, three to four hot inflation prints ahead, and no confirmed Fed chair.

ANDREW: Okay. What are we watching.

AVA: Ceasefire durability and the tolling mechanism. Jacob on Geopolitical Cousins gives Iran a two to four week leverage window before the international flotilla and alternative pipeline capex start eroding their position. Three violations already cited, Lebanon as a potential escalation trigger. The headline to watch is not "ceasefire" — it's whether ships are paying and transiting.

ANDREW: Kevin Worsh confirmation. Clarida's modal case from RenMac is Worsh chairs the June FOMC meeting. Steve Pavle apparently takes the over on May fifteenth. That sets the clock on when the Fed can credibly respond to any of this.

AVA: Software earnings. Q2 is the first data point on whether EPS actually rolls over to confirm what the multiple compression is already suggesting. That's when the private credit link becomes visible or doesn't.

ANDREW: Project Glass Wing. A hundred-day sandboxing window on Mythos ends around mid-July. That's when we find out if the cybersecurity capability is what Anthropic claims or what Chamath calls "fear-based marketing." The Chinese model parity window — Kimi K2 cited as roughly six months behind — runs on almost exactly the same clock.

AVA: And the IPO pipeline. SpaceX targeting June, OpenAI and Anthropic targeting Q4 but with question marks. Any IPO pricing below the eight-hundred-thirty-billion-dollar private valuation creates cascading write-downs across the exact private capital ecosystem Waxman was describing.

ANDREW: A lot of dominos, is what we're saying.

AVA: A lot of dominos. Arranged very close together.

ANDREW: That's Morning Signal for Saturday April eleventh. I'm Andrew.

AVA: I'm Ava. Go drink your coffee. You earned it.

ANDREW: Or maybe something stronger.

AVA: Or something stronger. We'll be back Monday. Take care of yourselves.