Bullpen.News From Edwin Dorsey
Daily recap Feedback

Thursday, September 3, 2026

From X

@HaydenCapital

Fred Liu on X

AppLovin - Interview with Xiaochuan (architect of $APP's Axon engine). Thought this was the most in-depth public conversation on AppLovin's business to date. (P.S. Youtube has great subtitle translation) Axon & Technical Philosophy - Axon was built by just 5 people in 3 months - AppLovin was the first to implement models with prediction windows exceeding 7 days - Deliberately avoids developing its own LLMs, preferring to use the best available models on the market — building LLMs in-house wouldn't generate the best ROI E-Commerce Expansion - Decision to enter e-commerce made in May 2024, with the product launching later that year; original team was only 10 people - Entering e-commerce was effectively building a new model from scratch - The gaming and e-commerce algorithms share nothing except infrastructure and accumulated organizational experience Core Competency & Culture - AppLovin's edge isn't industry-specific — it's rooted in how the company approaches problems (the culture itself), which he believes positions them to tackle challenges well beyond gaming and e-commerce - His relish for being called an "underdog" comes through clearly throughout the interview Social Media Ambitions - Already working on building a next-generation social media platform - Strategic logic: Meta started with captive organic traffic and built an ad platform on top; AppLovin is doing the reverse - starting with the ad platform and building toward organic traffic - Owning that organic traffic, if achieved, would be a significant advantage Short Seller Response - When the short reports were published, Xiaochuan personally reviewed the systems and code to assess whether the claims had merit Hiring Philosophy - After 2023, overhauled hiring: moved to paying among the highest salaries in Silicon Valley, and shifted focus away from seasoned veterans toward people with fewer than 2 years of experience - Core belief: experience is overrated, the capacity to learn is underrated; breakthroughs come from willingness to abandon convention - "You can't operate like a large company if you don't have a large company's resources" - Most prominent AI researchers are overpriced — high visibility and intense competition drives salaries beyond their marginal value - Targets talent at the "periphery of the spotlight": technically strong but undervalued by the market, and still carrying the underdog mentality that well-known names often lack (i.e. "Moneyball") - Entire engineering org runs on fewer than 100 people https://t.co/vXaWUZGkZt

Investor Letters

All letters

Microsoft ($MSFT) - Deep Dive

Microsoft’s AI data-center buildout lifted annual CapEx from $5.5 billion in FY2014 to $115.9 billion in FY2026, resetting free-cash-flow…

MSFT $3.8T
Bearhold Research 18 clicks

QXO Inc.

QXO is building a North American construction-products distributor through Beacon, Kodiak and TopBuild, targeting $18.1 billion of combined…

QXO $14B
SixSigmaCapital: Navigating the Financial Markets 1 click

Late August 2026 Random Ramblings

Higher rates could compress equity multiples and raise AI data center lease costs, while UWMC and Cogent raise questions about CEO-driven…

UWMC $2.2B
Yet Another Value Podcast 26 clicks

UK Primer: The Equity Revival

UK equities trade at decades-wide discounts to developed-market peers as improving sentiment, capital reform and foreign M&A support a…

ARM $269B
Aurelion Research 1 click

Publications we follow

Every publication on this site, hand-picked.

69 publications

Newsletters 53

YouTube 4

Blogs 11

Seeking Alpha 14 of 108

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.