Dimitry Nakhla | Babylon Capital®
RT @DimitryNakhla: 𝐂𝐡𝐫𝐢𝐬 𝐇𝐨𝐡𝐧, 𝐒𝐮𝐩𝐞𝐫 𝐂𝐨𝐦𝐩𝐚𝐧𝐢𝐞𝐬 & 𝐖𝐡𝐲 𝐆𝐫𝐨𝐰𝐭𝐡 𝐈𝐬𝐧’𝐭 𝐖𝐡𝐚𝐭 𝐌𝐨𝐬𝐭 𝐈𝐧𝐯𝐞𝐬𝐭𝐨𝐫𝐬 𝐓𝐡𝐢𝐧𝐤:

“Growth can come from two forms — price and volume… Most companies don’t have pricing power… But there is a special group of super companies that can price above inflation. And that’s, as Buffett taught, the test of whether you have the moat.

If you’re asking about volume growth… I may have low volume growth but a lot of pricing growth — that’s actually more important because of the leveraged effect… there’s no cost associated with it.”
___

𝐓𝐡𝐞 𝐥𝐞𝐬𝐬𝐨𝐧: 𝘕𝘰𝘵 𝘢𝘭𝘭 𝘨𝘳𝘰𝘸𝘵𝘩 𝘪𝘴 𝘤𝘳𝘦𝘢𝘵𝘦𝘥 𝘦𝘲𝘶𝘢𝘭. 𝙑𝙤𝙡𝙪𝙢𝙚-𝙙𝙧𝙞𝙫𝙚𝙣 𝙜𝙧𝙤𝙬𝙩𝙝 𝘰𝘧𝘵𝘦𝘯 𝘳𝘦𝘲𝘶𝘪𝘳𝘦𝘴 𝘤𝘢𝘱𝘪𝘵𝘢𝘭, 𝘧𝘢𝘤𝘦𝘴 𝘤𝘰𝘮𝘱𝘦𝘵𝘪𝘵𝘪𝘰𝘯, 𝘢𝘯𝘥 𝘵𝘺𝘱𝘪𝘤𝘢𝘭𝘭𝘺 𝘤𝘢𝘳𝘳𝘪𝘦𝘴 𝘮𝘦𝘢𝘯𝘪𝘯𝘨𝘧𝘶𝘭 𝘪𝘯𝘤𝘳𝘦𝘮𝘦𝘯𝘵𝘢𝘭 𝘤𝘰𝘴𝘵𝘴. 𝙋𝙧𝙞𝙘𝙞𝙣𝙜-𝙙𝙧𝙞𝙫𝙚𝙣 𝙜𝙧𝙤𝙬𝙩𝙝— 𝘸𝘩𝘦𝘯 𝘴𝘶𝘱𝘱𝘰𝘳𝘵𝘦𝘥 𝘣𝘺 𝘥𝘶𝘳𝘢𝘣𝘭𝘦 𝘤𝘰𝘮𝘱𝘦𝘵𝘪𝘵𝘪𝘷𝘦 𝘢𝘥𝘷𝘢𝘯𝘵𝘢𝘨𝘦𝘴 — 𝘣𝘦𝘩𝘢𝘷𝘦𝘴 𝘷𝘦𝘳𝘺 𝘥𝘪𝘧𝘧𝘦𝘳𝘦𝘯𝘵𝘭𝘺.
___

𝘼𝙣𝙙 𝙩𝙝𝙞𝙨 𝙞𝙨 𝙬𝙝𝙚𝙧𝙚 𝙞𝙣𝙘𝙧𝙚𝙢𝙚𝙣𝙩𝙖𝙡 𝙤𝙥𝙚𝙧𝙖𝙩𝙞𝙣𝙜 𝙢𝙖𝙧𝙜𝙞𝙣𝙨 𝙗𝙚𝙘𝙤𝙢𝙚 𝙘𝙧𝙞𝙩𝙞𝙘𝙖𝙡. 𝙄𝙣𝙘𝙧𝙚𝙢𝙚𝙣𝙩𝙖𝙡 𝙢𝙖𝙧𝙜𝙞𝙣 𝙨𝙞𝙢𝙥𝙡𝙮 𝙖𝙨𝙠𝙨:

For each new $1 of revenue, how much drops to operating profit?

Companies with genuine pricing power often exhibit:

• Higher incremental margins
• Stronger profit flow-through
• Minimal incremental cost

Because price increases largely bypass the cost structure.

𝘞𝘩𝘦𝘯 𝘪𝘯𝘤𝘳𝘦𝘮𝘦𝘯𝘵𝘢𝘭 𝘮𝘢𝘳𝘨𝘪𝘯𝘴 𝘢𝘳𝘦 𝘩𝘪𝘨𝘩, 𝘦𝘷𝘦𝘯 𝘮𝘰𝘥𝘦𝘴𝘵 𝘳𝘦𝘷𝘦𝘯𝘶𝘦 𝘨𝘳𝘰𝘸𝘵𝘩 𝘤𝘢𝘯 𝘵𝘳𝘢𝘯𝘴𝘭𝘢𝘵𝘦 𝘪𝘯𝘵𝘰 𝘰𝘶𝘵𝘴𝘪𝘻𝘦𝘥 𝘱𝘳𝘰𝘧𝘪𝘵 𝘨𝘳𝘰𝘸𝘵𝘩. 𝘛𝘩𝘪𝘴 𝘪𝘴 𝘵𝘩𝘦 𝘩𝘪𝘥𝘥𝘦𝘯 𝘦𝘯𝘨𝘪𝘯𝘦 𝘣𝘦𝘩𝘪𝘯𝘥 𝘮𝘢𝘯𝘺 𝘦𝘹𝘤𝘦𝘱𝘵𝘪𝘰𝘯𝘢𝘭 𝘤𝘰𝘮𝘱𝘰𝘶𝘯𝘥𝘦𝘳𝘴.
___

𝙀𝙭𝙖𝙢𝙥𝙡𝙚𝙨 𝙤𝙛 𝙨𝙪𝙥𝙚𝙧 𝙘𝙤𝙢𝙥𝙖𝙣𝙞𝙚𝙨 𝙩𝙝𝙖𝙩 𝙚𝙭𝙝𝙞𝙗𝙞𝙩 𝙩𝙝𝙚𝙨𝙚 𝙘𝙝𝙖𝙧𝙖𝙘𝙩𝙚𝙧𝙞𝙨𝙩𝙞𝙘𝙨:

• $FICO
• $ASML
• $NVDA
• $GE
• $TDG
• $MA
• $SPGI
• $MCO

Different industries. Similar underlying economics:

Durable moats + pricing power + strong incremental margins.
___

Video: In Good Company | Norges Bank Investment Management (05/14/2025)
tweet

DAIR.AI
RT @dair_ai: // Efficient Evolution of Web Agents //

Web agents waste a lot of compute on cyclic reasoning loops and unproductive exploration.

This new research introduces WebClipper, a framework that models web agent search processes as state graphs and prunes them into minimal directed acyclic graphs (DAGs).

The result: ~20% reduction in tool-call rounds while maintaining or improving accuracy.

They also introduce F-AE Score, a metric that evaluates the balance between accuracy and efficiency in agent trajectories.

Training agents on refined, pruned trajectories helps them develop more streamlined reasoning patterns from the start. Efficiency in agentic systems isn't just about faster models; it's really about eliminating wasted steps. This could significantly reduce costs as well.

Paper: https://t.co/GnvRt0VDq1
Learn to build effective AI agents in our academy: https://t.co/LRnpZN7L4c
tweet

Brady Long
I spoke to my Mac for 30 seconds.

90 minutes of work just... happened.

Emails sent. Docs written. Everything filed.

I made coffee while Lemon finished my day. https://t.co/zsGStTDu5g
tweet

DAIR.AI
RT @omarsar0: Managing rules for coding agents is a headache.

Claude Code, Cursor, Copilot... each uses its own standards.

Outdated rules derail coding agents, as we all know.

@QodoAI just shipped a rule system built on continuous learning.

It auto-discovers standards from your codebase and PR history, manages them centrally (dedup, conflict detection, severity levels), and gives you analytics to prove they're working.

It moves away from config files to a living, automated standards system.

The best part is that coding standards are now enforced in every PR.
tweet

DAIR.AI
A paper worth paying close attention to.

It presents Lossless Context Management (LCM), which reframes how agents handle long contexts.

It outperforms Claude Code on long-context tasks.

Recursive Language Models give the model full autonomy to write its own memory scripts. LCM takes that power back, handing it to a deterministic engine that compresses old messages into a hierarchical DAG while keeping lossless pointers to every original. Less expressive in theory, far more reliable in practice.

The results:

Their agent (Volt, on Opus 4.6) beats Claude Code at *every* context length from 32K to 1M tokens on the OOLONG benchmark. +29.2 points average improvement versus Claude Code's +24.7. The gap widens at longer contexts.

The implication is one we keep relearning from software engineering history: how you manage what the model sees may matter more than giving the model tools to manage it itself. Every agent framework shipping with "let the model figure it out" memory strategies may be building on the wrong abstraction entirely.

Paper: https://t.co/LtqS7pzmP4
Learn to build effective AI agents in our academy: https://t.co/LRnpZN7L4c
tweet

Dimitry Nakhla | Babylon Capital®
Pat Dorsey Q4 25’ 13F

Top 5 holdings: $ASML $DHR $AER $META $BKNG

Top Buys: $ASML $LYV

Top Sales: $GOOG
___

Source: Dataroma https://t.co/0goczNq4th
tweet

Javier Blas
The Iranian read-out sounds upbeat:

“We were able to reach broad agreement on a set of guiding principles, based on which we will move forward and begin working on the text of a potential agreement.” — Iran's Foreign Minister Abbas Araghchi

OIL MARKET: The 2nd round of US-Iran talks has concluded, and Iranian media says there would be a 3rd round of negotiations in the “near future” after both sides consult with their respective governments. - Javier Blas
tweet

Bourbon Capital
Chris Hohn still buying more $SPGI

Chris Hohn - TCI Fund Management Q4 2025 https://t.co/8r8I9llt5e
tweet

Moon Dev
be careful of openclaw!!

i lost $242,328 before my 6th openclaw was launched

grateful to have learned

I really dont want you to make the same mistake https://t.co/YWNKoTGjMb
tweet

God of Prompt
RT @godofprompt: 84% of developers use AI coding tools daily.
25% of new startups ship codebases that are almost entirely AI generated.

But here's the stat nobody talks about: AI assisted developers produce 3-4x more code... and 10x more security issues.

Your vibe coded app isn't broken. It's unfinished.

Here's the gap nobody's solving (and why it matters now): 🧵
tweet