Bullpen.News From Edwin Dorsey
Submit an Article

Friday, September 4, 2026

From X

@BrokenMoats 2 clicks

Broken Moats on X

$SNOW conference call is the bear case for application software companies. Their data infrastructure combined with coco and cowork on top increasingly makes SNOW a key player in data migrations, and data consolidation for their customers and a key hub to switch between models means SNOW has a seat at the table to be one of the most important and internal pieces of an AI stack....and as they become the permissions data holder with governance and api/mpc to internal and external partners, they can actually become the system of record across applications. This would allow you to use tools like cocoa inside to contextualize in human language what you want built, or even what data/answers you want and how you want that visualized.... Compare this to a Salesforce that is essentially becoming a database company silo to their specific domain by opening up their data to Claude to be the work engine. Eventually companies will increasingly want to stop interfacing with multiple applications to query and want to UI and query across a singular interface (originally thought to be an LLM like Claude, but with specific training and different inference costs / intelligence across models its more like the player presiding over the biggest pool of company data (snowflake, databricks) could become that central payer - less model lock, more flexibly, and central access point for models, data, queries and ui across functions. May not relegate a salesforce obsolete but significantly destroys their value when people access from Snow or even from an LLM like Claude were they lose the ui interface, and behavioral association and slight shared/learned network effects and become a database. Increases switching costs overtime and certainly destroys pricing (agent force consumption model is a step in the modern pricing setup, but start based models are dead, its not an if for most applications without highly defensible and proprietary data) $PEGA, $APPN, $PATH, $CRM, $PAYC are first pass through som of the more exposed names
@Mike10947310 3 clicks

Mike on X

During the Iran dip initially there was a lot of behind the scenes AI progress. And so when there was finally relief — in the form of the initial detente -> MOU — things totally ripped. I feel like with these insane $NVDA and $AVGO guides we are almost in this spot again. Hedge funds have delevered, momo has been historically crushed, Warsh has said a lot of hawkish things, and the next FOMC the market expects a hike. Obviously the big overhang is yields. But it feels to me like for all the discussion of guidance/rates/the yield curve… really the bond market just wants the war to end or at least be paused. I am not so sure that the US has no power to at least have things significantly deescalate when it wants (maybe into midterms). It’s like, we can “lose” a war, but Iran can still recognize that with time its leverage will decrease, its economy will be destroyed further… like it seems a bit naive just to think that because the war is going badly, Iran has no desire to see it wind down. This latest tit for tat for example hasn’t been that crazy. So IF you think that we get some sort of detente, which is really what I think yields need — are we not kind of a coiled spring here and running back a similar setup to when things started ripping?

Investor Letters

All letters

Publications we follow

Every publication on this site, hand-picked.

69 publications
Nominate a publication

X accounts we follow

Every account on this site, hand-picked.

55 accounts
Nominate an account

Sectors

Ordered by how often our editor reaches for them.