Offshore
Photo
Javier Blas
The oil ministers of Iran and Russia met today.
Contrary to popular belief, Moscow and Tehran are now bitter rivals in the oil market as the size of the black market for crude shrinks. Both compete to supply China.
(My earlier @Opinion take is below)
https://t.co/lZ94rzxzKz
tweet
The oil ministers of Iran and Russia met today.
Contrary to popular belief, Moscow and Tehran are now bitter rivals in the oil market as the size of the black market for crude shrinks. Both compete to supply China.
(My earlier @Opinion take is below)
https://t.co/lZ94rzxzKz
tweet
Offshore
Video
God of Prompt
Voice-to-action is where the agent layer is heading.
Most people underestimate how much productivity leaks through input friction. Typing, tab switching, copy pasting between apps. None of that is actual work. It's overhead.
The teams solving that friction layer will own the next wave of AI productivity.
Same core principle we've built around for years: the quality of your input determines the quality of your output.
Worth watching.
tweet
Voice-to-action is where the agent layer is heading.
Most people underestimate how much productivity leaks through input friction. Typing, tab switching, copy pasting between apps. None of that is actual work. It's overhead.
The teams solving that friction layer will own the next wave of AI productivity.
Same core principle we've built around for years: the quality of your input determines the quality of your output.
Worth watching.
Think it. Say it. Done.
The average person spends 3 hours typing + switches 1,000 tabs per day.
That ends today.
Meet Lemon: The first voice-to-action AI agent that turns your voice commands into finished tasks.
RT + Comment "Lemon" to get free access for 30 days.
(must be following so I can DM you) tweet
Offshore
Video
Chips & SaaS
RT @StockSavvyShay: Evercore says $NVDA Vera Rubin timeline has been pulled forward by 3โ6 months.
China export bans freed up supplier capacity, opening the door to shipments as early as end-Q2 2026. https://t.co/ruYBsSPhIW
tweet
RT @StockSavvyShay: Evercore says $NVDA Vera Rubin timeline has been pulled forward by 3โ6 months.
China export bans freed up supplier capacity, opening the door to shipments as early as end-Q2 2026. https://t.co/ruYBsSPhIW
tweet
Offshore
Photo
Benjamin Hernandez๐
Semi Sector: The Structural Support
While $MASI jumps on the Danaher acquisition, Big Tech is holding the index floor. Semiconductors remain the preferred vehicle for institutional capital flowing back into growth today.
$NVDA $MSFT $AMD $MU $PLTR
Watching the breakout. ๐ป๐ฅ https://t.co/qnOpuEmWJO
tweet
Semi Sector: The Structural Support
While $MASI jumps on the Danaher acquisition, Big Tech is holding the index floor. Semiconductors remain the preferred vehicle for institutional capital flowing back into growth today.
$NVDA $MSFT $AMD $MU $PLTR
Watching the breakout. ๐ป๐ฅ https://t.co/qnOpuEmWJO
tweet
Offshore
Photo
DAIR.AI
Training tool-use agents with RL requires diverse, executable environments.
But these environments barely exist.
This new research introduces Agent World Model (AWM), a fully synthetic pipeline that generates executable agentic environments at scale. Starting from high-level scenario descriptions, it synthesizes database schemas, tool interfaces exposed via MCP, and verification code, all backed by SQL databases for consistent state transitions.
It presents 1,000 unique environments spanning finance, travel, retail, social media, and more, with 35,062 tools and 10,000 tasks paired with verification code. Each environment supports parallel isolated instances for large-scale RL.
Scalable, executable environment synthesis is the missing piece for agentic RL. AWM demonstrates that agents trained exclusively in synthetic environments can generalize to out-of-distribution real-world benchmarks.
Paper: https://t.co/Hp2PpGsBzw
Learn to build effective AI agents in our academy: https://t.co/LRnpZN7L4c
tweet
Training tool-use agents with RL requires diverse, executable environments.
But these environments barely exist.
This new research introduces Agent World Model (AWM), a fully synthetic pipeline that generates executable agentic environments at scale. Starting from high-level scenario descriptions, it synthesizes database schemas, tool interfaces exposed via MCP, and verification code, all backed by SQL databases for consistent state transitions.
It presents 1,000 unique environments spanning finance, travel, retail, social media, and more, with 35,062 tools and 10,000 tasks paired with verification code. Each environment supports parallel isolated instances for large-scale RL.
Scalable, executable environment synthesis is the missing piece for agentic RL. AWM demonstrates that agents trained exclusively in synthetic environments can generalize to out-of-distribution real-world benchmarks.
Paper: https://t.co/Hp2PpGsBzw
Learn to build effective AI agents in our academy: https://t.co/LRnpZN7L4c
tweet
The Transcript
RT @TheTranscript_: $TRV CEO: AI is cutting renewal underwriting time by 30%+ in Personal Insurance
"our renewal underwriting platform leverages generative AI to consolidate data into summaries of relevant, actionable information for our underwriters to evaluate, with early results showing more than a 30% reduction in average handle time. The net result is that our underwriters focus their efforts on decisions most likely to improve profitability and do so more efficiently
tweet
RT @TheTranscript_: $TRV CEO: AI is cutting renewal underwriting time by 30%+ in Personal Insurance
"our renewal underwriting platform leverages generative AI to consolidate data into summaries of relevant, actionable information for our underwriters to evaluate, with early results showing more than a 30% reduction in average handle time. The net result is that our underwriters focus their efforts on decisions most likely to improve profitability and do so more efficiently
tweet
Offshore
Video
Startup Archive
Doug Leone: โIf youโre desperate, itโs a great assetโ
The Sequoia partner is asked what he looks for in entrepreneurs. He responds:
โI think entrepreneursโlike investorsโcome in different flavors. And I will tell you, as I told my kids, that if youโre desperate, itโs a great asset. If you have too many choices in life, it clouds your thinking. When you have only one way to go and thatโs forward, itโs very easy. You just go, go, go. Failure is truly not an option.โ
Doug and Sequoia look for people whoโve done quirky things and taken risks rather than โfollowed the same tracksโ as everyone else. They also look for domain expertise and people who are solving their own problem.
For example, when Jan Koum had a need for privacy and low cost messaging, he built WhatsApp.
โOr something as simple as Zappos,โ Doug continues. โThe founders of Zappos couldnโt find a pair of shoes! We look for people who are trying to solve problems that they themselves have.โ
Video source: @StanfordGSB (2014)
tweet
Doug Leone: โIf youโre desperate, itโs a great assetโ
The Sequoia partner is asked what he looks for in entrepreneurs. He responds:
โI think entrepreneursโlike investorsโcome in different flavors. And I will tell you, as I told my kids, that if youโre desperate, itโs a great asset. If you have too many choices in life, it clouds your thinking. When you have only one way to go and thatโs forward, itโs very easy. You just go, go, go. Failure is truly not an option.โ
Doug and Sequoia look for people whoโve done quirky things and taken risks rather than โfollowed the same tracksโ as everyone else. They also look for domain expertise and people who are solving their own problem.
For example, when Jan Koum had a need for privacy and low cost messaging, he built WhatsApp.
โOr something as simple as Zappos,โ Doug continues. โThe founders of Zappos couldnโt find a pair of shoes! We look for people who are trying to solve problems that they themselves have.โ
Video source: @StanfordGSB (2014)
tweet
Offshore
Video
Dimitry Nakhla | Babylon Capitalยฎ
RT @DimitryNakhla: ๐๐ก๐ซ๐ข๐ฌ ๐๐จ๐ก๐ง, ๐๐ฎ๐ฉ๐๐ซ ๐๐จ๐ฆ๐ฉ๐๐ง๐ข๐๐ฌ & ๐๐ก๐ฒ ๐๐ซ๐จ๐ฐ๐ญ๐ก ๐๐ฌ๐งโ๐ญ ๐๐ก๐๐ญ ๐๐จ๐ฌ๐ญ ๐๐ง๐ฏ๐๐ฌ๐ญ๐จ๐ซ๐ฌ ๐๐ก๐ข๐ง๐ค:
โGrowth can come from two forms โ price and volumeโฆ Most companies donโt have pricing powerโฆ But there is a special group of super companies that can price above inflation. And thatโs, as Buffett taught, the test of whether you have the moat.
If youโre asking about volume growthโฆ I may have low volume growth but a lot of pricing growth โ thatโs actually more important because of the leveraged effectโฆ thereโs no cost associated with it.โ
___
๐๐ก๐ ๐ฅ๐๐ฌ๐ฌ๐จ๐ง: ๐๐ฐ๐ต ๐ข๐ญ๐ญ ๐จ๐ณ๐ฐ๐ธ๐ต๐ฉ ๐ช๐ด ๐ค๐ณ๐ฆ๐ข๐ต๐ฆ๐ฅ ๐ฆ๐ฒ๐ถ๐ข๐ญ. ๐๐ค๐ก๐ช๐ข๐-๐๐ง๐๐ซ๐๐ฃ ๐๐ง๐ค๐ฌ๐ฉ๐ ๐ฐ๐ง๐ต๐ฆ๐ฏ ๐ณ๐ฆ๐ฒ๐ถ๐ช๐ณ๐ฆ๐ด ๐ค๐ข๐ฑ๐ช๐ต๐ข๐ญ, ๐ง๐ข๐ค๐ฆ๐ด ๐ค๐ฐ๐ฎ๐ฑ๐ฆ๐ต๐ช๐ต๐ช๐ฐ๐ฏ, ๐ข๐ฏ๐ฅ ๐ต๐บ๐ฑ๐ช๐ค๐ข๐ญ๐ญ๐บ ๐ค๐ข๐ณ๐ณ๐ช๐ฆ๐ด ๐ฎ๐ฆ๐ข๐ฏ๐ช๐ฏ๐จ๐ง๐ถ๐ญ ๐ช๐ฏ๐ค๐ณ๐ฆ๐ฎ๐ฆ๐ฏ๐ต๐ข๐ญ ๐ค๐ฐ๐ด๐ต๐ด. ๐๐ง๐๐๐๐ฃ๐-๐๐ง๐๐ซ๐๐ฃ ๐๐ง๐ค๐ฌ๐ฉ๐โ ๐ธ๐ฉ๐ฆ๐ฏ ๐ด๐ถ๐ฑ๐ฑ๐ฐ๐ณ๐ต๐ฆ๐ฅ ๐ฃ๐บ ๐ฅ๐ถ๐ณ๐ข๐ฃ๐ญ๐ฆ ๐ค๐ฐ๐ฎ๐ฑ๐ฆ๐ต๐ช๐ต๐ช๐ท๐ฆ ๐ข๐ฅ๐ท๐ข๐ฏ๐ต๐ข๐จ๐ฆ๐ด โ ๐ฃ๐ฆ๐ฉ๐ข๐ท๐ฆ๐ด ๐ท๐ฆ๐ณ๐บ ๐ฅ๐ช๐ง๐ง๐ฆ๐ณ๐ฆ๐ฏ๐ต๐ญ๐บ.
___
๐ผ๐ฃ๐ ๐ฉ๐๐๐จ ๐๐จ ๐ฌ๐๐๐ง๐ ๐๐ฃ๐๐ง๐๐ข๐๐ฃ๐ฉ๐๐ก ๐ค๐ฅ๐๐ง๐๐ฉ๐๐ฃ๐ ๐ข๐๐ง๐๐๐ฃ๐จ ๐๐๐๐ค๐ข๐ ๐๐ง๐๐ฉ๐๐๐๐ก. ๐๐ฃ๐๐ง๐๐ข๐๐ฃ๐ฉ๐๐ก ๐ข๐๐ง๐๐๐ฃ ๐จ๐๐ข๐ฅ๐ก๐ฎ ๐๐จ๐ ๐จ:
For each new $1 of revenue, how much drops to operating profit?
Companies with genuine pricing power often exhibit:
โข Higher incremental margins
โข Stronger profit flow-through
โข Minimal incremental cost
Because price increases largely bypass the cost structure.
๐๐ฉ๐ฆ๐ฏ ๐ช๐ฏ๐ค๐ณ๐ฆ๐ฎ๐ฆ๐ฏ๐ต๐ข๐ญ ๐ฎ๐ข๐ณ๐จ๐ช๐ฏ๐ด ๐ข๐ณ๐ฆ ๐ฉ๐ช๐จ๐ฉ, ๐ฆ๐ท๐ฆ๐ฏ ๐ฎ๐ฐ๐ฅ๐ฆ๐ด๐ต ๐ณ๐ฆ๐ท๐ฆ๐ฏ๐ถ๐ฆ ๐จ๐ณ๐ฐ๐ธ๐ต๐ฉ ๐ค๐ข๐ฏ ๐ต๐ณ๐ข๐ฏ๐ด๐ญ๐ข๐ต๐ฆ ๐ช๐ฏ๐ต๐ฐ ๐ฐ๐ถ๐ต๐ด๐ช๐ป๐ฆ๐ฅ ๐ฑ๐ณ๐ฐ๐ง๐ช๐ต ๐จ๐ณ๐ฐ๐ธ๐ต๐ฉ. ๐๐ฉ๐ช๐ด ๐ช๐ด ๐ต๐ฉ๐ฆ ๐ฉ๐ช๐ฅ๐ฅ๐ฆ๐ฏ ๐ฆ๐ฏ๐จ๐ช๐ฏ๐ฆ ๐ฃ๐ฆ๐ฉ๐ช๐ฏ๐ฅ ๐ฎ๐ข๐ฏ๐บ ๐ฆ๐น๐ค๐ฆ๐ฑ๐ต๐ช๐ฐ๐ฏ๐ข๐ญ ๐ค๐ฐ๐ฎ๐ฑ๐ฐ๐ถ๐ฏ๐ฅ๐ฆ๐ณ๐ด.
___
๐๐ญ๐๐ข๐ฅ๐ก๐๐จ ๐ค๐ ๐จ๐ช๐ฅ๐๐ง ๐๐ค๐ข๐ฅ๐๐ฃ๐๐๐จ ๐ฉ๐๐๐ฉ ๐๐ญ๐๐๐๐๐ฉ ๐ฉ๐๐๐จ๐ ๐๐๐๐ง๐๐๐ฉ๐๐ง๐๐จ๐ฉ๐๐๐จ:
โข $FICO
โข $ASML
โข $NVDA
โข $GE
โข $TDG
โข $MA
โข $SPGI
โข $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
RT @DimitryNakhla: ๐๐ก๐ซ๐ข๐ฌ ๐๐จ๐ก๐ง, ๐๐ฎ๐ฉ๐๐ซ ๐๐จ๐ฆ๐ฉ๐๐ง๐ข๐๐ฌ & ๐๐ก๐ฒ ๐๐ซ๐จ๐ฐ๐ญ๐ก ๐๐ฌ๐งโ๐ญ ๐๐ก๐๐ญ ๐๐จ๐ฌ๐ญ ๐๐ง๐ฏ๐๐ฌ๐ญ๐จ๐ซ๐ฌ ๐๐ก๐ข๐ง๐ค:
โGrowth can come from two forms โ price and volumeโฆ Most companies donโt have pricing powerโฆ But there is a special group of super companies that can price above inflation. And thatโs, as Buffett taught, the test of whether you have the moat.
If youโre asking about volume growthโฆ I may have low volume growth but a lot of pricing growth โ thatโs actually more important because of the leveraged effectโฆ thereโs no cost associated with it.โ
___
๐๐ก๐ ๐ฅ๐๐ฌ๐ฌ๐จ๐ง: ๐๐ฐ๐ต ๐ข๐ญ๐ญ ๐จ๐ณ๐ฐ๐ธ๐ต๐ฉ ๐ช๐ด ๐ค๐ณ๐ฆ๐ข๐ต๐ฆ๐ฅ ๐ฆ๐ฒ๐ถ๐ข๐ญ. ๐๐ค๐ก๐ช๐ข๐-๐๐ง๐๐ซ๐๐ฃ ๐๐ง๐ค๐ฌ๐ฉ๐ ๐ฐ๐ง๐ต๐ฆ๐ฏ ๐ณ๐ฆ๐ฒ๐ถ๐ช๐ณ๐ฆ๐ด ๐ค๐ข๐ฑ๐ช๐ต๐ข๐ญ, ๐ง๐ข๐ค๐ฆ๐ด ๐ค๐ฐ๐ฎ๐ฑ๐ฆ๐ต๐ช๐ต๐ช๐ฐ๐ฏ, ๐ข๐ฏ๐ฅ ๐ต๐บ๐ฑ๐ช๐ค๐ข๐ญ๐ญ๐บ ๐ค๐ข๐ณ๐ณ๐ช๐ฆ๐ด ๐ฎ๐ฆ๐ข๐ฏ๐ช๐ฏ๐จ๐ง๐ถ๐ญ ๐ช๐ฏ๐ค๐ณ๐ฆ๐ฎ๐ฆ๐ฏ๐ต๐ข๐ญ ๐ค๐ฐ๐ด๐ต๐ด. ๐๐ง๐๐๐๐ฃ๐-๐๐ง๐๐ซ๐๐ฃ ๐๐ง๐ค๐ฌ๐ฉ๐โ ๐ธ๐ฉ๐ฆ๐ฏ ๐ด๐ถ๐ฑ๐ฑ๐ฐ๐ณ๐ต๐ฆ๐ฅ ๐ฃ๐บ ๐ฅ๐ถ๐ณ๐ข๐ฃ๐ญ๐ฆ ๐ค๐ฐ๐ฎ๐ฑ๐ฆ๐ต๐ช๐ต๐ช๐ท๐ฆ ๐ข๐ฅ๐ท๐ข๐ฏ๐ต๐ข๐จ๐ฆ๐ด โ ๐ฃ๐ฆ๐ฉ๐ข๐ท๐ฆ๐ด ๐ท๐ฆ๐ณ๐บ ๐ฅ๐ช๐ง๐ง๐ฆ๐ณ๐ฆ๐ฏ๐ต๐ญ๐บ.
___
๐ผ๐ฃ๐ ๐ฉ๐๐๐จ ๐๐จ ๐ฌ๐๐๐ง๐ ๐๐ฃ๐๐ง๐๐ข๐๐ฃ๐ฉ๐๐ก ๐ค๐ฅ๐๐ง๐๐ฉ๐๐ฃ๐ ๐ข๐๐ง๐๐๐ฃ๐จ ๐๐๐๐ค๐ข๐ ๐๐ง๐๐ฉ๐๐๐๐ก. ๐๐ฃ๐๐ง๐๐ข๐๐ฃ๐ฉ๐๐ก ๐ข๐๐ง๐๐๐ฃ ๐จ๐๐ข๐ฅ๐ก๐ฎ ๐๐จ๐ ๐จ:
For each new $1 of revenue, how much drops to operating profit?
Companies with genuine pricing power often exhibit:
โข Higher incremental margins
โข Stronger profit flow-through
โข Minimal incremental cost
Because price increases largely bypass the cost structure.
๐๐ฉ๐ฆ๐ฏ ๐ช๐ฏ๐ค๐ณ๐ฆ๐ฎ๐ฆ๐ฏ๐ต๐ข๐ญ ๐ฎ๐ข๐ณ๐จ๐ช๐ฏ๐ด ๐ข๐ณ๐ฆ ๐ฉ๐ช๐จ๐ฉ, ๐ฆ๐ท๐ฆ๐ฏ ๐ฎ๐ฐ๐ฅ๐ฆ๐ด๐ต ๐ณ๐ฆ๐ท๐ฆ๐ฏ๐ถ๐ฆ ๐จ๐ณ๐ฐ๐ธ๐ต๐ฉ ๐ค๐ข๐ฏ ๐ต๐ณ๐ข๐ฏ๐ด๐ญ๐ข๐ต๐ฆ ๐ช๐ฏ๐ต๐ฐ ๐ฐ๐ถ๐ต๐ด๐ช๐ป๐ฆ๐ฅ ๐ฑ๐ณ๐ฐ๐ง๐ช๐ต ๐จ๐ณ๐ฐ๐ธ๐ต๐ฉ. ๐๐ฉ๐ช๐ด ๐ช๐ด ๐ต๐ฉ๐ฆ ๐ฉ๐ช๐ฅ๐ฅ๐ฆ๐ฏ ๐ฆ๐ฏ๐จ๐ช๐ฏ๐ฆ ๐ฃ๐ฆ๐ฉ๐ช๐ฏ๐ฅ ๐ฎ๐ข๐ฏ๐บ ๐ฆ๐น๐ค๐ฆ๐ฑ๐ต๐ช๐ฐ๐ฏ๐ข๐ญ ๐ค๐ฐ๐ฎ๐ฑ๐ฐ๐ถ๐ฏ๐ฅ๐ฆ๐ณ๐ด.
___
๐๐ญ๐๐ข๐ฅ๐ก๐๐จ ๐ค๐ ๐จ๐ช๐ฅ๐๐ง ๐๐ค๐ข๐ฅ๐๐ฃ๐๐๐จ ๐ฉ๐๐๐ฉ ๐๐ญ๐๐๐๐๐ฉ ๐ฉ๐๐๐จ๐ ๐๐๐๐ง๐๐๐ฉ๐๐ง๐๐จ๐ฉ๐๐๐จ:
โข $FICO
โข $ASML
โข $NVDA
โข $GE
โข $TDG
โข $MA
โข $SPGI
โข $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
Offshore
Photo
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
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