Index Company
Meta
19 CÈ pieces since 1 May 2026: 1 deal, 11 intel items, 7 essays. Newest first. All companies →
-
Singapore priced platform negligence at S$10M per breach; WhatsApp and Carousell have until January 2027
New Codes of Practice require messaging, social and e-commerce platforms to verify advertiser identity against government records, block unlicensed financial ads, and gate group adds. Breaches carry penalties up to S$10 million, with compliance due 31 January 2027.
-
Open weights arrive with a gatekeeper: Transformers v5.15 routes kernels through approved publishers
The release adds Meta's 30B Muse Glimmer alongside Granite-SWA, A.X-K1 and K2, and Cosmos3 Edge. Maintainers signalled the kernels package will become required for torch installs, sourced only from publishers Hugging Face approves.
-
Beijing can unwind a Singapore company's exit: Manus AI leaves Meta, $2B deal reversed
China's National Development and Reform Commission prohibited Meta's $2 billion purchase, which had closed in December 2025. Manus operates independently again and deletes data for users who registered on or after 29 December 2025.
-
When frontier models fit on one GPU, who needs the datacentre? Meta opens 30B Muse Glimmer
Meta released Muse Glimmer, a 30B open-weights multimodal model tuned for agentic tool use and coding, sized to run on a single consumer GPU. Zuckerberg says the weights of the larger Muse Spark 1.2 follow.
-
Sovereign AI in Southeast Asia (running on Hangzhou’s open weights)
Google says close to seven in ten Southeast Asian prompts now arrive in a local language, and the region has its own models to match; read the model cards and the sovereign part turns out to be the last layer.
-
Docs were written for humans: Meta's Astryx ships an agent-readable manifest, 9,000 stars in three weeks
Meta open-sourced Astryx, the React design system behind Facebook, Instagram, and Threads, under an MIT licence. Its CLI manifest exposes every command, flag, and response type as JSON, giving coding agents a specification instead of prose documentation.
-
Frontier agents become a metered utility: Meta's Muse Spark 1.1 opens a million-token context to developers
Meta Superintelligence Labs released Muse Spark 1.1 on July 9, a multimodal reasoning model built for agentic tasks, tool use, and coding, with a one-million-token context and a public-preview Meta Model API for developers.
-
Beijing weighs walling off the open weights that seeded Asia: curbs floated for Qwen, Doubao, and GLM-5.2
Reuters reported on July 7 that Chinese ministries met Alibaba, ByteDance, and Z.ai about limiting overseas access to advanced AI models, including open-weight releases. Officials floated bars on public release and treating model theft as a national security crime. Nothing is decided.
-
Grok 4.5 cut coding-AI prices by roughly 80 percent. Where to point it first.
xAI, OpenAI, and Meta all cut inference prices in the same week, and the cheapest coding model is now roughly twice as willing to make things up as the one it replaced.
-
What an AI agent can actually run in an SME back office right now
Two frontier models shipped in five weeks and the real product is admin, not code. The workflows an agent finishes alone are the ones that never leave the ledger; the ones that break are the ones that touch a person.
-
OpenClaw's local-first AI assistant framework crosses 380,000 GitHub stars
The self-hosted framework routes a personal AI agent across WhatsApp, Telegram, Slack, and Discord from a local gateway, reaching one of open source's fastest star counts on a privacy-first, cloud-independent design.
-
Meta's Business Agent turns WhatsApp into an always-on salesperson for merchants
The AI agent answers product questions, vets buyers, and closes orders inside WhatsApp, Messenger, and Instagram chats, with a Shopee integration aimed squarely at Southeast Asia's chat-commerce habits.
-
Southeast Asia's marketplaces raise seller fees as growth turns to monetisation
Shopee added a 5 percent technical-support fee across Singapore, Malaysia, Thailand and Vietnam this year, and TikTok Shop lifted Vietnam commissions to 12.5 percent, pushing total effective take rates toward 20 to 25 percent of post-discount sales.
-
The four-person company that does thirty people’s work (and rents the difference)
The same subsidised AI operators are warned not to lean on is, from the founder’s chair, the thing that lets four people do the work of thirty; in a region that has always starved origin, that moves the binding constraint.
-
The cheap AI on your desk is priced below cost
Four giants are losing money on every token to win the market, the bill for the buildout runs to US$5.5 trillion, and Singapore’s own forecasters have started naming the morning the subsidy stops.
-
OpenClaw passes 380,000 stars to become GitHub's most-starred project
Peter Steinberger's local-first AI assistant crossed 380,000 stars and 79,500 forks, overtaking React in roughly sixty days. It runs on your own devices and answers through WhatsApp, Telegram, Slack and Discord rather than a hosted chat box.
-
Zyphra releases ZAYA1-8B sparse model trained from scratch on AMD Instinct hardware under Apache 2.0
Zyphra's ZAYA1-8B model uses a sparse routing architecture with 8 billion total parameters but only 760 million active per token. Trained entirely on AMD Instinct hardware without NVIDIA dependency, it ships under Apache 2.0 license, demonstrating decentralised training infrastructure.
-
The 1.4 million deleted accounts that won’t fix the thing they were deleted for
The June 4 disruption removed a million accounts and froze a few million dollars, and the durable casualty, the one no warrant reaches, is that everyone in a regional chat now half-assumes the other side is running a script.
-
The year the Singapore middle stopped working.
Beijing blocked the Meta-Manus acquisition this month. The architecture of cross-border AI deals built around Singapore neutrality is being read again, by founders who built around it.