If you missed it, here’s what a summary recap of VC night, recordings and clips follow.
Why this exists
Conqueror opened with the eighteen months nobody heard from him.
The short version: after a health event he went back and did something he’d been threatening since 2021 — he turned all of Valhalla’s history into a dataset. Every position, every P&L, every argument about which strategy was right. What came out:
- Theta plays are profitable. They’re just unpopular. A dedicated theta server would be a dead server.
- Profitability tracks forum usage. Every spike on the trading floor lines up with a period when Catalysts or the forum was alive — when people were writing hypotheses, arguing entries, doing actual research. The eight months after monetization, when Catalysts was at its peak, was the greenest stretch across the most members in the server’s history.
- When the forums die, the community drifts into SPY. Most SPY positions in the server’s history are red overall. That’s not opinion, that’s the dataset.
So the thing that was pitched as the “Valhalla Community Trading System” in 2021, and the “crowdsourced quant” in 2022 after RSX, is what he spent the year and a half building. Three parts were covered:
Nexus — the boring part
A commercial-scale ingestion pipeline for basically everything market-related:
- Paid commercial newswires, plus every free wire
- Every SEC filing, with a dedicated reader per filing type (8-K, 10-K, earnings, etc.), all exhibits processed
- Change tracking on filings. Language like going concern only matters the first time it appears — companies keep filing the same boilerplate for years. Nexus diffs the current filing against the previous one and walks backward until it finds the first appearance. It also flags when language disappears, which is just as big a deal.
- FDA, FRED economic data, Congressional trades (ingested, not yet exposed)
- Relevance filtering. The ambulance-chaser lawyer press releases are ~gone. What you see when you scroll is tagged high/medium and is mostly things that actually matter.
His words: “This was ascension. If we could only do that.” This is the single page of only what matters we’ve been asking for since 2021.
Workspace + Meta Calendar
- Workspace is the thinkorswim grid, deployed in our software. Every panel is swappable, every panel reacts when you click into a ticker, and the whole layout lives in the URL — so you build the view you want and it’s yours.
- Meta Calendar merges everything that’s normally scattered across five sites: earnings, economic events, lockup expirations, FDA PDUFA dates, and Heimdall signals, on one calendar. “They’re all calendars. I don’t see the purpose of going to five of them.”
- Community tab shows positions and commentary together — the thing Catalysts was always supposed to be and never quite was.
Heimdall — the actual point
Everything above would have been enough. It’s not what he spent the time on.
Arcs. Type a ticker and you get an arc — the whole story, not a moment. Catalysts was built around the moment a stock had a reason to move; the only way to do better is to find the reason before it lands, and the only way to do that is to follow the story. An arc tracks a storyline through anticipation → event → ramification → confirmation → exhaustion, and tells you which phase it’s in.
The exhaustion phase is the part he emphasized most. The dataset showed we consistently lose money by overstaying trades. Heimdall makes an educated call on when it’s over, so someone is finally telling you to move on.
Evidence crosses tickers. An arc on UFG will pull in a sector partner’s earnings, a customer’s filing, a competitor’s guidance — whatever Heimdall decides is pertinent. When Dell reported, 16 independent arcs were watching that earnings print for margin signals and applying the result to their own storylines.
Scale. Tens of thousands of arcs, three types: ticker, cluster, macro. Watching the entire market at once. It wakes up on new events, on momentum, on random ticks, and on dates it schedules itself — the UFG arc has already booked wake-ups for November 12 and 15 because it expects a dilutive issuance around then. The furthest-out thing it’s currently tracking is three years away.
Projections. Each arc projects three orders out — this happens, then this, then this — with a probability the event occurs and a separate probability it actually moves the stock (because plenty of events land and do nothing). And every projection carries explicit invalidation conditions: this is wrong if the company commits to shareholder approval; this is wrong if they secure non-dilutive debt. It tells you how it’s wrong before it’s wrong.
Signals. Arcs exist for everything; a signal is when Heimdall decides there’s an opportunity and speaks up. ~15 a day. Each one has:
- a thesis in plain English
- “dies if” conditions — the price or event that kills the trade (on the Dell call the dies-if level held as the exact low of the post-earnings dip; buying the test would have been green the whole way)
- supersede/invalidate — it wakes back up and closes its own calls. Example from the same day: a signal built on the Intel Ohio fab leasing report was invalidated within hours when Intel denied it.
It’s not one LLM call. Ten-stage reasoning pipeline, four of which are LLM. The back end is hand-built; a modified version of the AMC-era short calculation is in there. It was not designed to be good at short squeezes — there’s very little of that in the training set — and it turned out to be good at them anyway, and better still at calling tops, because it tracks every company that has said it has shares to sell.
Calls he walked through
- UFG — flagged a Cayman Islands governance exemption in a routine filing that set up an unannounced dilutive issuance without NASDAQ shareholder approval. “We never would have caught that. I never would have.”
- ASST — 30% short float distorted by a non-cash Bitcoin markdown; the rest of the earnings were fine. Heimdall didn’t call it on earnings day — it woke up a week later on a random tick, saw the price hadn’t moved, and called the squeeze around $12.50. Ran to $27. “This was the one where it clicked that this was going to work.”
- NAVN — the squeeze call that survived the FOMC sell-off and tried to run into the close.
- GME, PANW, OKTA — green every day since signal. “We never had this many winners in a month in the entire history of the service.”
- LASE — flagged an ATM offering with volume too thin to absorb it. Down ~22% since.
- DELL — Generational wealth call. Dies-if entry, and the cross-arc margin-concern watch described above.
- Boeing — the honest one. Called on the Turkish Airlines resolution, then FOMC dumped the market and BA went with it. Conqueror admitted on the call that his first reaction was “it thinks it sold off because the thesis is invalid” — then read further and the signal already said if price reclaims 208, still on. Heimdall knew FOMC was today (a macro arc was watching it); the two just aren’t wired together yet. That’s a roadmap item, not a signal problem.
- RETO — not a Heimdall call. Ran ~1000% on zero news; the momentum scanner flagged it at 7,000x volume. Different tool, different job.
- Walmart — a bad signal, shown on purpose. A Barron’s retrospective was read as news and swayed the reasoning, already fixed at ingestion.
Also: positions are back, and better than the UI suggests. The position system now takes natural language ($GME bought 4 calls at X) instead of slash commands, and supports orders and complex options. The Bifrost positions UI is old and is next on the fix list.
FAQ — what got asked, what he answered
Is it technical? Is it watching the 200 MA?
No. Technicals aren’t fed to it. It has price levels and price action, and it pattern-matches shape — so it lands on things that look like technicals sometimes, but nothing traditional is in the input.
Does it catch sympathy plays — SMCI off Dell, Sandisk off Micron?
Yes, through cross-arc evidence and clusters. Clusters aren’t exposed in the UI yet (“they’re dense — real fucking dense”), but that’s exactly where that lives. The DRAM/storage miss from last year is the design case: everyone was watching DRAM, storage pricing went vertical, nobody was looking. Heimdall would have been.
Inverse sector trades — software up when hardware’s down?
Same answer: clusters. In the meantime, look at the pertinent macro arc plus a couple of the tickers under it and you’ll see where it thinks the pressure is going.
Macro — would it have told me to buy USO puts on the Saudi pipeline news?
It never tells you to buy anything. Macro arcs exist and generate signals (the OPEN play was macro), but they’re currently chained to 13. When unchained, you’ll get macro arcs for events like that. ETFs likely aren’t signal-eligible yet.
Is there a short-interest threshold?
No. Short interest is one piece of evidence among many. It’s called 13% SI names that ran like they were 50%. “I think there’s something about short interest we fundamentally don’t understand.”
What’s the difference between Watching and Live?
Watching = the arc exists, with evidence and projections, but Heimdall hasn’t decided there’s an opportunity. Live = it has spoken. If you trade a specific ticker, the watching arc is still valuable — the research is all there, it just doesn’t see a trade. (He doesn’t love the names; they’ll change.)
Is there one place to see the best plays?
That’s the Live board. Everything in Watching is Heimdall’s own homework, exposed.
Will it notify in Discord?
Yes — the market-open and market-alert channels are coming back with signal tie-ins (e.g. which signals moved most overnight). There will not be a phone ping for every signal; that’s noise and people would tune it out. He’s considering asking Heimdall itself which signals are worth a notification.
What about big macro days like FOMC?
The signals held up (BA was the only real casualty and it was still valid). The fix is UI: a banner when a macro arc is watching something big, so you know to check your positions.
Mobile?
On the list. Post what’s hard to use in the Support section of this forum — that’s where bugs and feature requests are being tracked now.
How does this get paid for?
Heimdall burns about half of Valhalla’s monthly membership revenue per day, and it’s currently chained — throttled processes, cheaper models than it was designed for. So:
- $25 is honored for everyone here now, and for select people who contributed to the community over the years. Valhalla Tots and the Valkyries are comped.
- Next round for people re-joining/people in the server that weren’t members before: $35–40.
- External pricing is ~$199/month. That is not passed to anyone here. con_dad’s reaction: he heard “$99” and thought it was low.
- Server capped around 1,000 members. (ridn2lo: “Witnessing the Thanos snap in real time is a high I’ve been chasing ever since.”)
- The ~2,000 dormant members in the server will be pinged before external marketing starts.
How are you marketing it?
Traditional channels — AdWords, the dormant Reddit account, word of mouth. Valhalla gets marketed as the community with Bifrost. Bifrost gets marketed on its own, separate from the Viking LARP, so there’ll be people using it who are never in the Discord. The trick is marketing it without giving away the edge — “they’re scraping the internet in mass.”
There are similar-looking products on YouTube.
“Get inside them and they’re hollow real quick.” The hard part isn’t calling one stock; it’s tens of thousands of arcs staying coherent instead of descending into madness — early iterations did exactly that. To his knowledge the only things running at this scale are institutional.
“You’re a great dev. You’re a shitty salesman.” — con_dad, seconded by Mysterious.
Conqueror took it. “Anyone who knows me outside the server would tell you the exact reverse. This was a lot of learning.” He may hire for marketing.
What if you unchain the projections past three orders?
“It gets extremely grim. Every single iteration decides we’re all going to die.” (sunny: “Bank run tomorrow, 9 a.m.”)
What he asked of us
This is the phase. The platform’s up; what’s missing is the activity. Comment on arcs, post positions, file bugs and feature requests in Support. “If we had this platform in 2021 we would have been going nuts, and there’s no reason it can’t be the same now.”
His point on why Heimdall calls things so early: most of what it flags has a time horizon. It’s not a scalping tool. Calling anticipation-phase and squeezes early is deliberate — it’s trying to get people green on things farther out, because green is the only reason most people hold.
Next VC: Tomorrow 9/17, 9:00 AM pre-bell — live board walk, then the open.
Then: Morning Show Monday @ 8:00 AM
Then: VC Night again Wednesday @ 8:30 PM
Also in attendance: Dook, who was there the whole time while Conq lamented his absence. CCIV DM’d Duke and has been re-invited, Conq cheered. Machete Phil, Dustin, Asinite, and Holi’s G-wagon were all remembered somewhat fondly, may Dustin and Holi’s G-wagon rest in peace. Squishy is buying scratch tickets now and there was broad concern.