whowe

Finance × AI × Property
Building wealth through data & deals.
Dividend growth | Johor real estate | Pickleball athlete 🏓

[whowe @weehowe]

@VictoryOwai Manager doesnt decide anything. As an arsenal player if you keep feeding martinelli/trossard and they either pass back to you or lose the ball, you may become insecure and stop picking that route.

[whowe @weehowe]

@iPaulCanada 重点在脚步

[whowe @weehowe]

@Dekchaikumkom The dog chews the AirPods with the shoes

[whowe @weehowe]

@NousResearch Guys I haven’t installed the Dashboard nor the Mac app. Need some tips. When to use Dashboard when to use the app?

[whowe @weehowe]

https://t.co/zTV0OLvdGe

[whowe @weehowe]

@Teknium can we set (maybe up to 3) model/provider so we can switch easily while in telegram etc. eg: /model 1, /model 2

[whowe @weehowe]

Great News

[BFM News @NewsBFM]

Malaysia is on track to host family offices managing up to RM2 billion in assets at Forest City by end-2026, driven by rising interest from wealthy investors.

The Securities Commission said nine family offices have already been set up under the Single Family Office incentive scheme, managing about RM670 million in assets so far.

🧵1

[whowe @weehowe]

CLI

[HeyGen @HeyGen]

Your AI agent can now generate and ship videos.

HeyGen CLI is now live.

Run one command and your agent handles it all:
script → avatar creation → video → delivery

All from the terminal. Just your agent and the CLI.

RT + Comment “CLI” and we’ll DM API credits (must follow)

[HeyGen @HeyGen]

Your AI agent can now generate and ship videos.

HeyGen CLI is now live.

Run one command and your agent handles it all:
script → avatar creation → video → delivery

All from the terminal. Just your agent and the CLI.

RT + Comment “CLI” and we’ll DM API credits (must follow)

[whowe @weehowe]

🚨 BREAKTHROUGH: MSA (Memory Sparse Attention)

Someone just gave LLMs ACTUAL long-term memory.

Not RAG. Not bigger windows. Memory BUILT INTO attention itself.

Why everything else fails:
❌ RAG = open-book exam. Can't remember, only searches. Cross-doc reasoning? Dead.
❌ Linear attention = compressed memory. Gets blurry. Loses details.

MSA is different:
→ Native sparse attention, LINEAR complexity
→ 10x memory, SAME compute cost
→ Knows WHERE & WHEN info came from
→ Connects scattered clues into reasoning chains

The numbers are insane:
• 16K → 100M tokens, <9% accuracy drop
• 4B MSA BEATS 235B RAG systems
• 2x A800 GPUs = 100M token inference

Before: Genius with goldfish memory
After: Actually remembers

https://t.co/BFHAZgB4Fm (https://t.co/BFHAZgB4Fm)

Star it 🌟

#AI #LLM #MachineLearning #Attention #Memory

[whowe @weehowe]

Like + RT for Part 2:
"Which other sovereign funds are about to crack?"

#Gold #Thailand #PensionFund #IranWar #Qatar #LNG #Commodities #Trading #Finance

[whowe @weehowe]

🚨 WHO DUMPED $400M GOLD DURING ASIA HOURS?

The gold crash wasn't hedge funds.
It wasn't JP Morgan.

It was a SOVEREIGN PENSION FUND in distress.

Here's the smoking gun 👇

[whowe @weehowe]

Sources: Reuters, Bangkok Post, Bloomberg, AInvest

This is why gold dumped while oil surged.

Not inflation. Not Fed.

A Thai pension fund hitting its risk limit.

/END