Offshore
Photo
Michael Fritzell (Asian Century Stocks)
AeroEdge just hit JPY 4,000/share, making it an 8-bagger within a year. I no longer own shares. But water your flowers and cut your weeds, is my take-away. https://t.co/PlUqTZR3rW
tweet
AeroEdge just hit JPY 4,000/share, making it an 8-bagger within a year. I no longer own shares. But water your flowers and cut your weeds, is my take-away. https://t.co/PlUqTZR3rW
tweet
Offshore
Photo
Michael Fritzell (Asian Century Stocks)
RT @gvancomp: The forgotten lovechild of MSCI (FTSE) postponed the March review for Indo. https://t.co/i3bzNzqPWe
Random thoughts from my favourite indomie
Firstly picture this, I'm sitting on the airplane, next to the bathrooms where someone obviously had 1 too many nasi lemaks, on a turbulent plane ride, wearing a wife beater, and a whiskey soda resting on my oversized belly. And with that image in mind, do not take this as investment advice, my #s are off and my blood is full of msg from all the wonderful indomie I've consumed.
What a time for Indo ... Apa ini ya...
Patriot bonds at 2% (47/50 investors are Chinese, amusing isn't it?) these bonds are issued by the presidents office and not by the government ...
Nephew at the central bank
Danantara sucking up capital and not putting anything to work
Martabe taken (apparently they didn't even know it was foreign owned..seriously what phucwit forget to check? Btw apparently it will be returned shortly)
And now MSCI....
Goreng goreng
The goreng boys just went too wild
Goreng (stock pumping/manipulation) happens everywhere in the world
But not to the point where 6 of the 18 names in the MSCI EM are goreng
So they've brought this on themselves
And they did it so well...
Active funds, both local and international have been 8itching about this for years. The goreng boys have done a fantastic job at making MSCI look like an subpar Enron Auditor, and shafted large international capital, but MSCI woke up and decided to slap some sense into the Indo capital markets burst the super goreng game.
Indonesia has until May to placate the global overlords of passive money....
So now what
A few scape goats, the Heads of IDX and OJK stepped down.
Is that enough? Some more brokers will need to be taken down 1? 2? Or 50 jailed? (the fancy ones of course where they can still trade from, play padel and go home on weekends) But I still suspect a more visual number of heads will need to roll.
Now the current head of OJK is well known to be a swimsuit model in her younger days. Perhaps she has her own epstein black book of names to maintain some control and order...
What happens if Indo gets kicked out of EM?
That's USD 40 BN in passive capital that will have to leave. For a country with USD 110 BN in reserves...that's a pretty penny ...the Fx would drop 30-50%. Yields moonshot. Side note Vietnam gets shafted in the MSCI frontier as Indo would become 40% and Vietnam drops from 30 to 18%
What "solutions" have happened
Well they are going to require the free float minimum of 15%
Means ... the bankers are going to busy ... but domestic AM isn't large enough to absorb all of this.
Big boys and friends can simply set up vehicles via local AMs, "Private Equity or Hedge Funds" in Singapore and HK vehicles to ensure that on paper this threshold is met...but that's still a lot of capital required ...
If you're a big player...move fast before the available capital is dried up.
Or because this is a mess...
Two exchanges? A main board and a PoS board?
Exchange A. One for those meet the free float requirement
Exchange B. One for those that don't.
Exchange A can be the main board and considered part of regional and global indices
In theory these are the "better" companies and they'll see a bid in the shift from the goreng names shifted to Exchange B which will be sold down and those flows will go to Exchange A quality names.
Danantara has apparently started putting capital into the market. Let's see how much support that can give...
Danantara is also amusing...again a good idea. Odd execution. But can you imagine being a person on the investment team there providing recommendations? What if it doesn't pan out despite best intentions? Would you be fined/jailed under the next administration? (Look what they've done to the former Minister of Education whom actually tried to do something positive) I doubt any middle manager or j[...]
RT @gvancomp: The forgotten lovechild of MSCI (FTSE) postponed the March review for Indo. https://t.co/i3bzNzqPWe
Random thoughts from my favourite indomie
Firstly picture this, I'm sitting on the airplane, next to the bathrooms where someone obviously had 1 too many nasi lemaks, on a turbulent plane ride, wearing a wife beater, and a whiskey soda resting on my oversized belly. And with that image in mind, do not take this as investment advice, my #s are off and my blood is full of msg from all the wonderful indomie I've consumed.
What a time for Indo ... Apa ini ya...
Patriot bonds at 2% (47/50 investors are Chinese, amusing isn't it?) these bonds are issued by the presidents office and not by the government ...
Nephew at the central bank
Danantara sucking up capital and not putting anything to work
Martabe taken (apparently they didn't even know it was foreign owned..seriously what phucwit forget to check? Btw apparently it will be returned shortly)
And now MSCI....
Goreng goreng
The goreng boys just went too wild
Goreng (stock pumping/manipulation) happens everywhere in the world
But not to the point where 6 of the 18 names in the MSCI EM are goreng
So they've brought this on themselves
And they did it so well...
Active funds, both local and international have been 8itching about this for years. The goreng boys have done a fantastic job at making MSCI look like an subpar Enron Auditor, and shafted large international capital, but MSCI woke up and decided to slap some sense into the Indo capital markets burst the super goreng game.
Indonesia has until May to placate the global overlords of passive money....
So now what
A few scape goats, the Heads of IDX and OJK stepped down.
Is that enough? Some more brokers will need to be taken down 1? 2? Or 50 jailed? (the fancy ones of course where they can still trade from, play padel and go home on weekends) But I still suspect a more visual number of heads will need to roll.
Now the current head of OJK is well known to be a swimsuit model in her younger days. Perhaps she has her own epstein black book of names to maintain some control and order...
What happens if Indo gets kicked out of EM?
That's USD 40 BN in passive capital that will have to leave. For a country with USD 110 BN in reserves...that's a pretty penny ...the Fx would drop 30-50%. Yields moonshot. Side note Vietnam gets shafted in the MSCI frontier as Indo would become 40% and Vietnam drops from 30 to 18%
What "solutions" have happened
Well they are going to require the free float minimum of 15%
Means ... the bankers are going to busy ... but domestic AM isn't large enough to absorb all of this.
Big boys and friends can simply set up vehicles via local AMs, "Private Equity or Hedge Funds" in Singapore and HK vehicles to ensure that on paper this threshold is met...but that's still a lot of capital required ...
If you're a big player...move fast before the available capital is dried up.
Or because this is a mess...
Two exchanges? A main board and a PoS board?
Exchange A. One for those meet the free float requirement
Exchange B. One for those that don't.
Exchange A can be the main board and considered part of regional and global indices
In theory these are the "better" companies and they'll see a bid in the shift from the goreng names shifted to Exchange B which will be sold down and those flows will go to Exchange A quality names.
Danantara has apparently started putting capital into the market. Let's see how much support that can give...
Danantara is also amusing...again a good idea. Odd execution. But can you imagine being a person on the investment team there providing recommendations? What if it doesn't pan out despite best intentions? Would you be fined/jailed under the next administration? (Look what they've done to the former Minister of Education whom actually tried to do something positive) I doubt any middle manager or j[...]
Offshore
Michael Fritzell (Asian Century Stocks) RT @gvancomp: The forgotten lovechild of MSCI (FTSE) postponed the March review for Indo. https://t.co/i3bzNzqPWe Random thoughts from my favourite indomie Firstly picture this, I'm sitting on the airplane, next to…
unior would dare offer any investment ideas. If so get a second home ready in Singapore... - Pon Van Compernolle tweet
Offshore
Photo
Offshore
Photo
Michael Fritzell (Asian Century Stocks)
Google Trends seems to be broken, presumably due to traffic from generative AI bots. Any fix? https://t.co/pufXKRSvxV
tweet
Google Trends seems to be broken, presumably due to traffic from generative AI bots. Any fix? https://t.co/pufXKRSvxV
tweet
Offshore
Photo
Michael Fritzell (Asian Century Stocks)
RT @kos_data: 🇨🇳 Xi net approval rating in Europe:
Top fans:
🇷🇸 Serbia: 22
🇲🇩 Moldova: 15
🇬🇷 Greece: 8
Top haters:
🇸🇪 Sweden: -81
🇺🇦 Ukraine: -69
🇻🇦 Pope Leo XIV net approval rating in Europe:
Top fans:
🇭🇷 Croatia: 67
🇵🇱 Poland: 46
🇮🇪 Ireland: 44
Top haters:
🇹🇷 Turkey: -31
🇷🇸 Serbia: -19
🇮🇱 Netanyahu net approval rating in Europe:
Netanyahu only fans:
🇺🇦 Ukraine: 2
Top haters:
🇧🇦 Bosnia: -80
🇳🇴 Norway: -75
🇸🇪 Sweden: -71
🇮🇳 Modi net approval rating in Europe:
Top fans:
🇬🇷 Greece: 16
🇲🇩 Moldova: 12
🇷🇸 Serbia: 6
Top haters:
🇺🇦 Ukraine: -35
🇩🇰 Denmark: -34
🇱🇹 Lithuania: -32
tweet
RT @kos_data: 🇨🇳 Xi net approval rating in Europe:
Top fans:
🇷🇸 Serbia: 22
🇲🇩 Moldova: 15
🇬🇷 Greece: 8
Top haters:
🇸🇪 Sweden: -81
🇺🇦 Ukraine: -69
🇻🇦 Pope Leo XIV net approval rating in Europe:
Top fans:
🇭🇷 Croatia: 67
🇵🇱 Poland: 46
🇮🇪 Ireland: 44
Top haters:
🇹🇷 Turkey: -31
🇷🇸 Serbia: -19
🇮🇱 Netanyahu net approval rating in Europe:
Netanyahu only fans:
🇺🇦 Ukraine: 2
Top haters:
🇧🇦 Bosnia: -80
🇳🇴 Norway: -75
🇸🇪 Sweden: -71
🇮🇳 Modi net approval rating in Europe:
Top fans:
🇬🇷 Greece: 16
🇲🇩 Moldova: 12
🇷🇸 Serbia: 6
Top haters:
🇺🇦 Ukraine: -35
🇩🇰 Denmark: -34
🇱🇹 Lithuania: -32
🇺🇸 US President Trump net approval rating in Europe:
Top fans:
🇽🇰 Kosovo: 27
🇷🇴 Romania: 11
🇲🇩 Moldova: 10
🇲🇰 N.Macedonia: 2
Top haters:
🇩🇰 Denmark: -84
🇸🇪 Sweden: -80
🇳🇴 Norway: -79
(Poll: Gallup International) https://t.co/XCAXSStNBt tweet
Offshore
Photo
DAIR.AI
RT @dair_ai: What if you could get multi-agent performance from a single model?
Multi-agent debate systems are powerful. Multiple LLMs can critique each other's reasoning, catch errors, and converge on better answers.
However, the cost scales linearly with the number of agents. Five agents means 5x the compute. Twenty agents means 20x and so on.
But the intelligence gained from debate doesn't have to stay locked behind a compute wall.
This new research introduces AgentArk, a framework that distills the reasoning capabilities of multi-agent debate into a single LLM through trajectory extraction and targeted fine-tuning.
This work addresses an important problem: multi-agent systems are effective but expensive at inference time. AgentArk moves that cost to training time, letting a single model carry the reasoning depth of an entire agent team.
The key idea: run multi-agent debate offline to generate high-quality reasoning traces, then train a smaller model to internalize those patterns.
Five agents debate, one student learns.
AgentArk tests three distillation methods. RSFT uses supervised fine-tuning on correct trajectories. DA filters for diverse reasoning paths. PAD, their strongest method, preserves the full structure of multi-agent deliberation, capturing how agents verify intermediate steps and localize errors.
The results across 120 experiments:
> PAD achieves a 4.8% average gain over single-agent baselines, with in-domain improvements reaching up to 30%. On reasoning quality metrics,
> PAD scores highest in intermediate verification (4.07 vs 2.41 baseline) and reasoning coherence (3.96 vs 1.88 baseline).
>The distilled models also transfer: trained on math, they improve on TruthfulQA with ROUGE-L jumping from 0.613 to 0.657.
Scaling from Qwen3-32B teachers down to Qwen3-0.6B students, the framework holds up. Even sub-billion parameter models absorb meaningful reasoning improvements from multi-agent debate.
Paper: https://t.co/cyPTig221s
Learn to build effective AI agents in our academy: https://t.co/LRnpZN7L4c
tweet
RT @dair_ai: What if you could get multi-agent performance from a single model?
Multi-agent debate systems are powerful. Multiple LLMs can critique each other's reasoning, catch errors, and converge on better answers.
However, the cost scales linearly with the number of agents. Five agents means 5x the compute. Twenty agents means 20x and so on.
But the intelligence gained from debate doesn't have to stay locked behind a compute wall.
This new research introduces AgentArk, a framework that distills the reasoning capabilities of multi-agent debate into a single LLM through trajectory extraction and targeted fine-tuning.
This work addresses an important problem: multi-agent systems are effective but expensive at inference time. AgentArk moves that cost to training time, letting a single model carry the reasoning depth of an entire agent team.
The key idea: run multi-agent debate offline to generate high-quality reasoning traces, then train a smaller model to internalize those patterns.
Five agents debate, one student learns.
AgentArk tests three distillation methods. RSFT uses supervised fine-tuning on correct trajectories. DA filters for diverse reasoning paths. PAD, their strongest method, preserves the full structure of multi-agent deliberation, capturing how agents verify intermediate steps and localize errors.
The results across 120 experiments:
> PAD achieves a 4.8% average gain over single-agent baselines, with in-domain improvements reaching up to 30%. On reasoning quality metrics,
> PAD scores highest in intermediate verification (4.07 vs 2.41 baseline) and reasoning coherence (3.96 vs 1.88 baseline).
>The distilled models also transfer: trained on math, they improve on TruthfulQA with ROUGE-L jumping from 0.613 to 0.657.
Scaling from Qwen3-32B teachers down to Qwen3-0.6B students, the framework holds up. Even sub-billion parameter models absorb meaningful reasoning improvements from multi-agent debate.
Paper: https://t.co/cyPTig221s
Learn to build effective AI agents in our academy: https://t.co/LRnpZN7L4c
tweet
Offshore
Photo
DAIR.AI
RT @omarsar0: This is a great read if you are building complex applications with Claude Code and Codex.
Most AI coding agents can generate a frontend.
But building a real full-stack application is a completely different story.
The gap between generating a landing page and shipping a working app with a functional backend, database, and API layer remains wide.
Most coding agents default to mock data, fake endpoints, and frontend-only implementations.
But real-world web development requires all three layers working together.
This new research introduces FullStack-Agent, a multi-agent system designed for end-to-end full-stack web development with two key innovations: Development-Oriented Testing and Repository Back-Translation.
The system uses three specialized agents. A Planning Agent generates structured frontend and backend designs in JSON. A Backend Coding Agent implements server logic with a dedicated debugging tool that sends HTTP requests and validates responses. A Frontend Coding Agent builds the UI against real backend APIs with a tool that monitors terminal and browser console errors dynamically.
Development-Oriented Testing validates code during generation, not after. Each agent gets real-time execution feedback, catching integration failures as they happen rather than at the end of a long generation chain.
Repository Back-Translation solves the training data problem. An information-gathering agent reads real open-source repositories and extracts development patterns. A trajectory agent then reproduces those repositories from scratch given only the extracted plans, generating high-quality training examples grounded in real codebases.
The results on their FullStack-Bench (647 frontend, 604 backend, 389 database test cases): FullStack-Dev achieves 64.7% frontend accuracy, 77.8% backend accuracy, and 77.9% database accuracy. That's an 8.7%, 38.2%, and 15.9% improvement over the strongest baseline respectively.
After training a Qwen3-Coder-30B model on 2K crawled and 8K augmented trajectories, frontend accuracy improved by 9.7% and backend accuracy by 9.5% in just two rounds.
The bottleneck in AI-assisted web development isn't frontend generation. It's building functional backends and databases that actually work together. FullStack-Agent closes that gap with execution-grounded testing and real-world training data.
Paper: https://t.co/b2g041Pvrb
Learn to build effective AI agents in our academy: https://t.co/1e8RZKs4uX
tweet
RT @omarsar0: This is a great read if you are building complex applications with Claude Code and Codex.
Most AI coding agents can generate a frontend.
But building a real full-stack application is a completely different story.
The gap between generating a landing page and shipping a working app with a functional backend, database, and API layer remains wide.
Most coding agents default to mock data, fake endpoints, and frontend-only implementations.
But real-world web development requires all three layers working together.
This new research introduces FullStack-Agent, a multi-agent system designed for end-to-end full-stack web development with two key innovations: Development-Oriented Testing and Repository Back-Translation.
The system uses three specialized agents. A Planning Agent generates structured frontend and backend designs in JSON. A Backend Coding Agent implements server logic with a dedicated debugging tool that sends HTTP requests and validates responses. A Frontend Coding Agent builds the UI against real backend APIs with a tool that monitors terminal and browser console errors dynamically.
Development-Oriented Testing validates code during generation, not after. Each agent gets real-time execution feedback, catching integration failures as they happen rather than at the end of a long generation chain.
Repository Back-Translation solves the training data problem. An information-gathering agent reads real open-source repositories and extracts development patterns. A trajectory agent then reproduces those repositories from scratch given only the extracted plans, generating high-quality training examples grounded in real codebases.
The results on their FullStack-Bench (647 frontend, 604 backend, 389 database test cases): FullStack-Dev achieves 64.7% frontend accuracy, 77.8% backend accuracy, and 77.9% database accuracy. That's an 8.7%, 38.2%, and 15.9% improvement over the strongest baseline respectively.
After training a Qwen3-Coder-30B model on 2K crawled and 8K augmented trajectories, frontend accuracy improved by 9.7% and backend accuracy by 9.5% in just two rounds.
The bottleneck in AI-assisted web development isn't frontend generation. It's building functional backends and databases that actually work together. FullStack-Agent closes that gap with execution-grounded testing and real-world training data.
Paper: https://t.co/b2g041Pvrb
Learn to build effective AI agents in our academy: https://t.co/1e8RZKs4uX
tweet
Offshore
Photo
Michael Fritzell (Asian Century Stocks)
RT @ai: TSMC is going to manufacture advanced AI semiconductors in Japan. This is a big deal for supply chain diversification. For decades, cutting-edge chip fabrication was concentrated in Taiwan. Now TSMC is building serious capacity in Japan, Arizona, and Germany.
https://t.co/EHng9bVPd6
tweet
RT @ai: TSMC is going to manufacture advanced AI semiconductors in Japan. This is a big deal for supply chain diversification. For decades, cutting-edge chip fabrication was concentrated in Taiwan. Now TSMC is building serious capacity in Japan, Arizona, and Germany.
https://t.co/EHng9bVPd6
tweet