New Ask Claude or ChatGPT about any stock, using Bullpen. Bullpen in Claude or ChatGPT. 7 days free

Friday, October 2

Friday, October 2, 2026

Everything we published on this day.

67 stories — the front page that day →

The Week That Was

Treasury yields surged as the Fed’s expected hike shifted to December, while Micron beat guidance and forecast $61.5 billion in quarterly…

Fajasy on X

If ur reading a write-up and they don't include a valuation range, more often than not: 1) They're lying to themselves. E.g. only pitching bull case. "The next 10x." $GRAB a couple years ago. 2) They're lying to you, the reader. E.g. not focusing on bear case enough. "It's asymmetric"... but actual downside if couple thing flops is 60%+. $FLNC just 2-3 months ago. 3) They have no clue and prob conclude by saying "position as small speculative bet" to be in a win-win scenario to gain subs. $CCXI today... Multiple based valuation is the simple/lazy route and easier to defend, yet less practical. Absolute valuations like DCF or Penman REV forces you to think more through scenarios and the business itself, which is what's actually useful because it's more subjective.

Fajasy on X

The $VIX closed at 14.87 on September 25, while the S&P 500's $SPY realized volatility over the past month was 10.8. That 4.1-point gap is the premium option buyers pay over what the index has actually done, and it's the newest row in the volatility section of my macro dashboard. The other volatility rows (VIX, the MOVE index for Treasury volatility, the VIX term structure, and SKEW) all live inside the options market, while this one checks option prices against the index's actual moves. A wide gap means protection is expensive relative to recent moves. A negative gap means the market is moving more than options priced in, which is what a selloff in progress looks like. I added it because a VIX of 15 means something different when the S&P 500 is moving 10% annualized than when it's moving 20%. The gap is near its 5-year average of 3.3 after swinging from -1.2 in early July to 8.4 by late July, 0.6 in mid-August, and 6.7 on September 11. Stocks look calm next to bonds right now. The VIX sits at the 23rd percentile of its 5-year range, while the MOVE index jumped from 80.6 to 96.0 in the week to September 25.

Maius on X

Some $PSIX thoughts off the recent 8-K: The banking syndicate lifted committed capacity from $135m to $220m (+$85m), cut the SOFR margin from +2.60% to +1.80%, and pushed maturity from July 2027 to September 2029 (~26 months). Credit committees are structurally risk-averse. More exposure, less spread, and longer tenor only clears if they like the repayment case. So what are they underwriting that most equity investors still won’t touch? Two things the disclosures imply: 1) Weichai’s North American strategy is locked in. Change-of-control remains tied to Weichai keeping board control and ≥40% ownership (same economic threshold as before). The extension removes left-tail risk that Weichai pursues a different go-to-market path. 2) The extra capacity isn’t for today’s balance sheet. Net debt was essentially zero at Q2’26 (cash ~$70m, debt ~$73m). The new investor materials flag Prime Power as a deliverable for the first time—ranges that overlap $INIO’s data-center gas engines, a segment where $INIO recently highlighted as an area "Chinese players have historically struggled to break in" and hence command "100% market share in several datacenter power ranges". As Bezos put it: your margin is my opportunity. Have a great weekend, everybody!

Convequity on X

Whether AI can run security operations on its own comes down to one question: when an automated action is wrong, who is accountable for finding out what happened and fixing it? Our note borrows two levels from self-driving cars. At L2 the driver answers for mistakes. At L4 the car drives itself, so accountability shifts to the company behind the system. 1. At L2, a third-party AI agent mostly assists. It connects to a security platform over MCP, the protocol that lets an outside AI model pull data from a vendor's platform. It suggests next steps and drafts investigations, while the human and the security platform still hold the wheel. 2. At L4 or L5, the system acts with no human in the loop in the security operations centre. It quarantines a machine, kills a process, revokes access or blocks traffic. Any of those can be wrong. When one is, the customer needs a clear accountable party that will investigate and fix the problem. In our view $PANW can be that party. It is already inside the customer's systems, keeps audit trails, and its own products take those hard-to-reverse actions. AI model labs are poorly placed for that role. On our reading, their terms usually limit what they answer for, so when an outside agent acts wrongly, the customer is left to investigate and fix the problem alone. Outside agents compete most easily for the analyst's workspace, which is L2 work. That makes $PANW better placed than model labs to sell autonomy.

David Orr on X

Google $GOOG Key bear argument: Search is their largest business and likely faces a long term headwind instead of long term tailwind. At least probably, since they are doing a good job incoperating fast AI responses into search. Android (and iOS and Windows) are threatened. The work culture is weak (but seems to be improving). Key bull argument: Most of their other businesses are good/great and growing. A couple are big AI winners. TPUs in particular are a likely key AI ingredient if either/both inference and electricity efficiency matter a lot (I think yes to both). YouTube is a big AI winner. Cloud/compute biz wins from AI, deals are structured in a way with no downside for $GOOG (but plenty of downside for the companies locking into the contracts). Management are fantastic capital allocators.

junkbondinvestor on X

Cable One's ($CABO) unsecured debt is trading under 30 cents. Its 1L TL is under 80. This is a company that was ONCE valued at $13bn+ and now has a measly market cap of <$100mm. Part of the reason (besides the whole sector being in shambles)? It owes GTCR an est. $480mm for the rest of MBI, a cableco whose price was locked in at the TOP of the cable cycle. That payment was due yesterday. This morning, they announced a 1-week extension and disclosed that they're in advanced financing talks with GTCR, PC lenders, and "certain" existing lenders. Here's what I think it means for the existing debt as well as equity and what could happen next

Convequity on X

$NVDA is still the cheapest way to produce AI tokens when 150 or more users share one batch. When one user wants the fastest possible answer, the economics change. Our Cerebras note puts the rule in two parts. 1. NVIDIA is the cheapest token factory when 150+ concurrent users share one batch and each user gets a moderate speed. 2. To maximise speed for a single user, the batch has to shrink, so the system's total output falls. That loses money per token unless the faster answers are sold at a higher price. Cerebras ($CRBS) sits at the other end. Its wafer-scale chip gives fast answers to a single user and reaches its best efficiency at 4 users. Like a taxi, it is quick and direct, but it has fewer seats to spread the cost across. The two also differ in how much room each has left to improve. SRAM-based designs like Cerebras have already captured most of their performance upside. GPU fleets often run at low compute utilization today, and that gap is room for future gains and economies of scale, not a permanent cost. So fast inference only pays if customers will pay more for speed. Our view is that a sizeable portion of the market will pay a premium for speed.

Nike, Part Two: The Deep Dive

Nike’s second Latticework program examines the company and the global athletic footwear and apparel industry for professional investors.

Nike, Part One: The Story

Nike grew from a $1,000 shoe-distribution venture in 1964 into the largest sports brand, using athlete endorsements, product innovation and…