Bullpen.News From Edwin Dorsey
Get Daily Recap Feedback

Friday, September 25, 2026

Back to today

From X

@RagingVentures 3 clicks

Raging Capital Ventures on X

Caught a nice move in $META as they have executed nicely on Muse. However, I’m curious to see Muse’s retention stats. Muse’s agentic capabilities still seem like a bit of a novelty to me, and sometimes more of a hassle to use than a value-add. $META has also aggressively re-targeted me with ads on my feed based on my Muse activity. While impressive from a business monetization point of view, it makes me leery to connect my email and other sensitive accounts.
@RagingVentures 2 clicks

Raging Capital Ventures on X

Along with our great guest speakers (outlined below), our Ideas & Networking Conference next Thursday, October 1st in NYC will feature a few investment ideas from yours truly. Over the years, we’ve highlighted some good ones, including: 2023 — $EFXT below $6, now $23+; $FPH below $3, now $4.65 2024 — Alphawave Semiconductor, subsequently acquired by $QCOM at a 70% premium; $MRAM below $6, now $18; $WBD below $9, now $30 2025 — Short $ORCL at $285 Plus, many more ideas along the way. We’re down to just a handful of seats for next week. Click below to register: https://t.co/NZIcFTuLGW
@convequity 3 clicks

Convequity on X

We’ve been mega-bullish on $PLTR since late 2020. We expected it to come out on top of our SaaS Survival Scorecard. It did. Most software will sit under someone else’s AI window. Palantir is one of the few that can reasonably be that window — the place where work, decisions, and data meet. The edge is breadth plus depth. Breadth: hundreds of kinds of operational work on one substrate, not one app. Depth: it reaches down into the source data layer. A lot of the records still live in old ERP and CRM systems. Palantir doesn’t have to own those databases. It normalizes the mess, puts a shared map on top — objects, relationships, allowed actions — and that map becomes the ground truth. Think of SAP and Salesforce as the ground floor. Palantir is floor 1: where the business actually runs. That creates data gravity and speed. Once decisions live on floor 1, going around Palantir is the hard path. One thing that is not a proven Palantir product yet is refining models on the customer’s own missions — post-training and reinforcement learning on their data, with the customer keeping the weights. That is the NVIDIA Nemotron work. Until recently it was a partnership. Two weeks ago at AIPCon 11, NVIDIA showed it live in its own supply chain. That’s the first public install. Not a general customer rollout. A few months ago the stock was subdued even after strong results, on the fear that model labs would expand into Palantir’s territory. A working Nemotron collaboration takes some of that risk off the table. The risk now is different: Nemotron has to become a frontier-quality model. If it does, Palantir is the best — and probably the only — company positioned to deliver true sovereignty over enterprise intelligence. If it isn’t smart enough, post-training and RL on Palantir’s stack may still not be good enough for enterprises. Watch two windows for Nemotron progress. The nearer one is Q3 earnings commentary — Palantir usually reports that quarter in early November. Management may say more there. The other is the next AIPCon, Palantir’s customer case-study day, roughly every quarter - possibly early December. That’s where live customer proof will show up.

Investor Letters

All letters

In Conversation: NVIDIA

NVIDIA sees inference, agentic workloads and robotics broadening AI compute demand beyond hyperscalers, while high-speed memory shortages…

NVDA $5.5T
Aurelion Research 1 click

Publications we follow

Every publication on this site, hand-picked.

110 publications

Newsletters 93

YouTube 4

Blogs 12

Seeking Alpha 41 of 198

Nominate a publication

X accounts we follow

Every account on this site, hand-picked.

72 accounts
Nominate an account

Investor letters we follow

189 firms

Sectors

Ordered by how often our editor reaches for them.