ANDREW: Good morning, good morning — happy Sunday, April twelfth. I'm Andrew.

AVA: And I'm Ava. Andrew, I have to say, I've been up since six staring at my notes for today and I genuinely don't know where to start. There's a lot going on.

ANDREW: That's the polite way to put it. We've got an active war in the Middle East, oil above a hundred bucks a barrel, a Fed chair whose replacement is being held hostage by a DOJ investigation, and an AI company claiming it found a twenty-seven-year-old bug hiding in critical infrastructure. So... just a regular Sunday.

AVA: Just your average weekend. Let's get into it. You want to start with markets?

ANDREW: Yeah, let me kick it off because the setup here is genuinely strange. So over on Excess Returns — their April ten through twelve episode — the hosts are working through what they're calling a historically anomalous situation. You've got oil above a hundred dollars a barrel. Active conflict in Iran. And yet the VIX is sitting at nineteen.

AVA: Wait, nineteen? After everything that's happened?

ANDREW: Nineteen. Down from a peak of around thirty-six during the escalation. And here's the part that stopped me cold — Excess Returns flagged that the speed of that vol compression, from roughly the twenty-five level down to where we are now, has only one historical precedent in the data.

AVA: Which is...

ANDREW: 2007. The months before the Global Financial Crisis.

AVA: Okay that's... that's a sentence I did not need to hear on a Sunday morning.

ANDREW: Right. And they're not saying we're heading into a GFC replay — but they're saying the speed of the compression is worth taking seriously. The reason the market didn't freak out when oil broke a hundred is actually somewhat rational. The U.S. naval buildup in the region was visible for weeks beforehand, so professional money had already pre-positioned into the escalation. Excess Returns called it the "taco trade" — Trump Always Chickens Out — essentially the market pricing in a de-escalation even as the conflict was active.

AVA: So the ceasefire tweet gap-up was kind of... already expected?

ANDREW: Largely priced, yeah. And that gap-up actually created its own problem — the S&P gapped through the JP Morgan collar put strike around sixty-four seventy-five, which was marking the absolute March low. Once you blow through that level, you get a negative gamma rally, dealers buying as the market moves higher, and suddenly you're grinding toward six thousand one hundred.

AVA: Which is now resistance.

ANDREW: Heavy resistance. There was aggressive call selling after the gap-up — Excess Returns is pretty specific about this — and then even heavier resistance clusters at sixty-eight to sixty-nine hundred. The options market is essentially pricing in no upside. Single-stock options in mag-seven names are almost entirely put-heavy, essentially zero call positioning. Index-level options are call-weighted. The interpretation from Excess Returns is that the market is either correctly skeptical of a rally, or it's dangerously under-hedged for a scenario where something resolves favorably and you get a gamma squeeze on the way up.

AVA: So the asymmetry is actually skewed toward upside pain, not downside?

ANDREW: That's the nuanced read, yeah. Realized vol has normalized — they're running around twelve percent, which is roughly seventy-five basis points of daily S&P moves — but the spread between VIX and realized that was very wide in March has fully compressed. And look, we've got VIX expiration Wednesday, equity OpEx Friday. Excess Returns notes that historically two thirds of the time you see a trend reversal after OpEx. The market has rallied hard into this expiration, which to their mind suggests default softness afterward.

AVA: And then earnings season starts, which is its own wild card.

ANDREW: "Earnings and tweets" — that's literally how Excess Returns frames the remaining directional drivers. Both unknown. Okay, let me hand you the inflation picture because RenMac this week had Richard Clarida on and that conversation is worth digging into.

AVA: Yeah, so on RenMac — this was their April ten through twelve episode featuring Clarida — the overall read on inflation is actually more constructive than the headline CPI from April tenth might suggest. Headline was hot, but core was restrained. And Clarida walks through the underlying data: the Dallas Fed trimmed mean PCE is running at roughly half the pace of core PCE, which he takes as evidence that the disinflation process is "largely intact." Wage inflation adjusted for productivity is tracking toward target. Housing cost pressures are easing. The five-year, five-year forward inflation expectation is still pinned near two percent.

ANDREW: And the bond selloff we've seen recently — that's been real rates, not inflation expectations breaking out.

AVA: Exactly. Clarida makes that point explicitly on RenMac. The move in yields is being driven by real rates, not a re-pricing of inflation. Which is why he describes five-year TIPS as relatively attractive right now. The base case for Clarida is a first rate cut in the second half of twenty-twenty-six, probably March to June is what the market's pricing, but his conditional is important: that depends entirely on whether the Middle East energy shock is short-lived.

ANDREW: And if it's not short-lived?

AVA: Then we could be looking at twenty-twenty-seven for the first cut. And that's not a tail scenario — that's what he calls a "very, very significant wildcard." A hundred dollar oil that persists through the summer doesn't just delay cuts. It threatens the entire disinflation narrative that's been underneath equity multiples since twenty-twenty-four.

ANDREW: Which is the feedback loop that nobody's really pricing at VIX nineteen.

AVA: Nobody's pricing. And then you layer on the Fed Chair situation, which is... Andrew, honestly, this might be the most underappreciated story in the briefing.

ANDREW: Talk me through it because RenMac had some details here I hadn't seen elsewhere.

AVA: So on RenMac, Clarida walks through the Kevin Warsh confirmation picture. Warsh is Trump's nominee to replace Powell as Fed Chair. The modal case had been a confirmation around the June FOMC meeting. But Senator Tillis — Republican from North Carolina — has explicitly conditioned his vote on the DOJ dropping its criminal investigation of Jay Powell.

ANDREW: The DOJ is investigating the sitting Fed Chair.

AVA: That is the situation we are in, yes. And Acting AG Blanche is expected to take direction from the White House, but RenMac reports no movement yet. So you've got this extraordinary circumstance where monetary policy succession is being determined not by economics or Senate arithmetic, but by the status of a DOJ investigation.

ANDREW: And in the meantime Powell is essentially a lame duck.

AVA: A lame duck whose successor has a specific rate-cut philosophy that cannot yet be acted upon. RenMac actually pulls out the Eccles precedent — nineteen fifty-one, Treasury-Fed Accord — where a replaced Fed chair stayed on as a governor and brokered policy through coalition-building with career Fed staff. Because here's the thing most people miss: the FOMC Chair is elected annually by the committee itself. It doesn't automatically transfer to whoever the President appoints as Board Chair.

ANDREW: So Powell could theoretically remain influential even after Warsh is confirmed.

AVA: If he's kept on as a governor and not recess-appointment-blocked — yes. The complexity of this transition is not being priced. Markets are treating it as a binary Warsh confirmed or not confirmed. RenMac is saying there are several intermediate scenarios that are very much on the table.

ANDREW: That's a fat tail that doesn't look fat in current options pricing. Alright, let's shift gears because you've been sitting on the Anthropic stuff and I can see you're vibrating.

AVA: I am vibrating. Okay so on the All-In Podcast this week — their April ten through twelve episode — the main event is a disclosure from Anthropic about a model called Mythos that they have apparently chosen not to release publicly. And the reason they held it back is that during internal testing, the model discovered a twenty-seven-year-old vulnerability in OpenBSD. A sixteen-year-old bug in FFmpeg that had survived five million automated security scans. Linux kernel vulnerabilities. And crucially — it could chain three to five of these bugs together into novel exploit sequences.

ANDREW: Novel meaning previously unknown attack combinations.

AVA: Right. Not just finding bugs — synthesizing them into new weapons. And in response to this, Anthropic apparently worked with the government to stand up something called Project Glass Wing: a hundred-day coalition of forty companies, including Apple, Microsoft, Google, Amazon, JP Morgan, using Mythos offensively to find and patch dormant vulnerabilities before a hostile actor does.

ANDREW: Okay but I want to push back here because the All-In hosts were split on this, right? Sacks and Chamath were not buying the framing.

AVA: They were not. And look, it's a fair skeptical read. On All-In, Sacks and Chamath's argument is that Anthropic has a documented pattern of using fear as a go-to-market strategy. They point to the earlier "blackmail study" that got a lot of attention — required over two hundred prompts to produce any result and generated zero real-world incidents. And they go all the way back to twenty-nineteen, GPT-2, when OpenAI said the model was too dangerous to release and then released it six months later to basically no incident.

ANDREW: The boy who cried dangerous AI.

AVA: That's the Sacks framing, essentially. But here's where I land — and this is closer to where Brad and Jason Cal are on All-In — the logic is actually pretty sound. If Mythos is genuinely the best coding model in the world, it would naturally discover vulnerabilities first. That's not a marketing claim, that's just... what the best coding model would do.

ANDREW: And China isn't standing still.

AVA: All-In flags that Kimi K2 — that's China's leading model — is roughly six months behind on open-source benchmarks. Six months is not a lot of runway. And critically, that's the open-source benchmark comparison. What their closed-door capabilities look like is genuinely unknown. The non-zero probability that China already has Mythos-equivalent cyber tools is the scenario that keeps the national security framing from being pure theater.

ANDREW: What I found interesting is the absence of certain sectors from Project Glass Wing. Energy companies, nuclear reactor operators, aviation — not in the forty.

AVA: Which is a real contradiction if the threat is truly existential. Those are some of the most critical infrastructure sectors you could imagine. All-In flags it and doesn't fully resolve it. It's either a significant oversight or evidence that the threat is being somewhat selectively applied.

ANDREW: Let me jump in on something else from All-In because the OpenClaw situation is directly relevant to the business dynamics here, and I think it's getting lost under the Mythos headline. So there's a developer named Peter Steinberger who built a tool called OpenClaw — it's the number one project on GitHub right now — built on top of Claude's API. He was paying two hundred dollars a month for a subscription. On April eighth, Anthropic cut him off and forced him to pay-per-token at the API rate.

AVA: Which is... a lot more expensive.

ANDREW: Significantly more expensive if you're a power user. And on the same day — the same day — Anthropic launched their own first-party product called Claude Managed Agents, which is a direct competitor to what OpenClaw was doing.

AVA: That's a pretty aggressive move.

ANDREW: It's the AWS playbook. Watch what your best customers are building, then build it yourself at the infrastructure layer. All-In raises the antitrust angle: if Anthropic holds something like fifty to sixty percent of the coding token market and bundles its own agent at a flat rate while charging metered API to third-party competitors... that's not a frivolous bundling and price-dumping argument.

AVA: The counterargument is that the market being contested — coding tokens — is still only about five percent of the total AI coding market. It's not like this is the whole game.

ANDREW: Yet. The question is whether that five percent becomes fifty percent before regulators or open-source alternatives can constrain Anthropic's pricing power. And speaking of open-source alternatives — the BitTensor story from All-In is genuinely wild.

AVA: Ridges AI.

ANDREW: Ridges AI, running on BitTensor's Tau subnet sixty-two. A team of anonymous developers achieved roughly eighty percent of Claude Code's capability in forty-five days, spending about a million dollars in token rewards. And they're pricing it at twenty-nine dollars a month.

AVA: Versus two hundred for Claude Pro.

ANDREW: Versus two hundred. Now Brad on All-In pushes back — and he's right — that enterprise-grade, long-horizon code generation is not there yet. Fortune five hundred companies are not running payroll on a crypto-incentivized anonymous subnet. But Jason Cal's counter is also right: the tooling and agent layer is already migrating to open source. The mid-market tier — the Cursor users, the Claude Pro subscribers — is getting squeezed from below by Ridges and Venice, and from above by Mythos and whatever OpenAI's next model is.

AVA: Spud, apparently.

ANDREW: Spud. Which All-In describes as the first Blackwell-trained OpenAI model, expected imminently, with similar capabilities to Mythos.

AVA: Great name.

ANDREW: Really rolls off the tongue. And look, this compression maps directly to what Excess Returns is seeing in the equities. Software sector making new lows — Monday.com, Teams, Asana all flagged specifically. The market is already pricing the AI deflationary impulse into the stocks that built the previous software wave.

AVA: Let me bring in the macro angle on AI because RenMac actually has a fascinating disconnect here. Clarida and Warsh disagree on this in a way that matters for monetary policy. On RenMac, Clarida says AI's net inflationary versus deflationary effect over the next three to five years is a "much closer call" than people assume. Near term, AI capex is inflationary — construction, power, servers, Nvidia chips, all of that is old economy demand pulling forward. The productivity and deflationary benefits are deferred three to five years out.

ANDREW: And Warsh?

AVA: Warsh is on record saying AI is strongly disinflationary. Which matters because Warsh is the guy who, if confirmed, will be setting rate policy. If he's wrong about the disinflationary timeline — if Clarida is right that the payoff is deferred — and he cuts based on an AI productivity assumption that hasn't materialized yet, you could end up with a serious policy error.

ANDREW: And RenMac notes that this debate has already moved into the policy apparatus. Clarida mentions Treasury Secretary Bessent convening major banks specifically to discuss new AI models.

AVA: Right. This isn't just a tech sector conversation anymore. The capability threshold debate is inside the Treasury Department. If the Mythos disclosure is real and government is already coordinating with Anthropic, the timeline to formal AI regulation or deployment restrictions compresses — and that directly affects valuations across the entire agentic software layer.

ANDREW: Alright, let me bring it home with what we're watching. VIX expiration is Wednesday. Excess Returns has been pretty explicit: the market has rallied hard into this expiration, two thirds of the time you get a trend reversal post-OpEx, the default lean is softness heading into next week.

AVA: And then equity OpEx Friday. Smaller than March's quarterly in delta terms but Tesla specifically was flagged as a name to watch. Post-OpEx direction sets the tone into earnings season.

ANDREW: Q1 earnings — All-In and Excess Returns both land in the same place, which is basically... no conviction. "Earnings and tweets" as the two remaining directional drivers, both unknowable. Which is honest, if not exactly actionable.

AVA: On the geopolitical side, I'm watching the ceasefire situation closely because that's the hinge point for everything. If the peace-tweet holds and oil comes back below a hundred, Clarida's base case on rate cuts — second half of twenty-twenty-six — becomes more credible, equities have a reason to push through six thousand one hundred resistance, and the OpEx softness might be contained. If ceasefire fails and oil holds above a hundred through summer, you're looking at cuts pushed to twenty-twenty-seven, Warsh confirmation calculus changes again, and the soft landing thesis that's implicitly priced into these S&P levels gets directly challenged.

ANDREW: The Warsh confirmation deadline is the June FOMC meeting under the modal case. But Tillis's blockage has introduced what RenMac calls a "fat tail" extending into summer. And that DOJ investigation of Powell — no movement yet — is the random variable nobody has a model for.

AVA: And I'm watching Project Glass Wing's hundred-day clock. It started roughly April eighth. The first real public test of whether Mythos is national security reality or sophisticated marketing will be the initial disclosures of vulnerabilities found and patched. If forty major companies have been running an AI-powered audit of their codebases for a hundred days and the output is... fifteen minor bugs that would've been caught anyway... that tells you something. If the output is a list of critical infrastructure vulnerabilities that could have been weaponized — that tells you something very different.

ANDREW: The asymmetry in this market is interesting. Near term, you've got VIX at nineteen, calls cheap at the index, resistance at six thousand one hundred, and OpEx mechanics that historically favor softness. But the left tail — ceasefire failure, Warsh confirmation collapse, or a hostile actor making a Mythos-equivalent disclosure before we've patched our own infrastructure — none of those are remotely priced at current volatility levels.

AVA: The setup, as you'd say, favors caution into Friday.

ANDREW: As I would say. Alright, that is Morning Signal for Sunday April twelfth. We'll be back tomorrow as the week gets underway. VIX expiration Wednesday, OpEx Friday, keep an eye on oil and keep an eye on the ceasefire situation.

AVA: And read your vendor contracts carefully, because apparently your coding tool provider might launch a competing product on the same day they cut off your API access.

ANDREW: Cold-blooded. Thanks for listening, everyone. Stay sharp out there.

AVA: See you tomorrow.