The Prior Briefing

AI Product Development

Anthropic's Mythos Model Launches with Project Glasswing Security Coalition; Run-Rate Revenue Hits $30B

Anthropic launched its Mythos model alongside Project Glasswing, a security initiative involving over 40 major tech companies including Apple, Google, Microsoft, Cisco, and Broadcom. Anthropic is providing $100 million in Mythos usage credits to partners and donating $4 million to open-source security efforts so participants can scan and patch vulnerabilities in their own systems. Mythos is capable of identifying and exploiting zero-day vulnerabilities across every major OS and browser, including a 27-year-old OpenBSD bug; Anthropic has withheld general public access due to cybersecurity risk. Separately, Anthropic announced at HumanX conference that its annualized run-rate revenue surpassed $30 billion, up from $9 billion at end of 2025. The company is valued at $380 billion, and sentiment among Silicon Valley VCs has shifted sharply in its favor over OpenAI, valued at $852 billion.

Why it matters

As a Singapore-based product leader and investor, Anthropic's trajectory — tripling revenue in three months, shipping a frontier model with enterprise security partnerships, and preparing for IPO — makes it a credible platform bet for enterprise AI product builds and a compelling investment thesis relative to OpenAI's richer valuation. Project Glasswing's coalition also signals that API access to Mythos's capabilities may open to enterprise partners, which is directly relevant if you are evaluating foundational model providers for production systems.

Sources: Techdirt, Business Insider (first: Techdirt)

Alibaba's HappyHorse Video Model Tops Video Arena Benchmark, API Access Planned by April 30

Alibaba's HappyHorse 1.0, developed by its new Alibaba Token Hub unit led by CEO Eddie Wu, scored 1387 points on the Video Arena benchmark to rank first in both text-to-video and image-to-video categories, with API access expected by April 30.

Sources: Panda Daily, South China Morning Post (first: Panda Daily)

Meta AI Upgraded with Muse Spark Model, Benchmarked Against ChatGPT

Meta's upgraded Meta AI, powered by the new Muse Spark model with a multiagent approach, demonstrated location-aware and opinion-forward responses in direct testing against ChatGPT, reflecting its social media data roots.

Source: TechRadar

Google DeepMind and Google Cloud Deploy 3D Biomechanical AI for Team USA Winter Olympics

Google DeepMind and Google Cloud built a real-time 3D pose estimation system for U.S. Winter Olympians that plots 63 joints per frame using TPUs statically provisioned on Vertex AI to eliminate cold-start latency during competition bursts.

Source: Google Cloud Blog

Foundation Models & Llm Infrastructure

Anthropic's Mythos Model Launches With $100M Security Initiative and Surging Revenue

Anthropic launched its new frontier model, Mythos, during the HumanX conference in San Francisco this week, announcing it cannot yet be released to the general public due to cybersecurity risks. Mythos is capable of identifying and exploiting zero-day vulnerabilities across all major operating systems and browsers, including a now-patched 27-year-old bug in OpenBSD. Alongside the model, Anthropic announced Project Glasswing, a coalition of over 40 major tech companies including Apple, Google, Microsoft, Cisco, and Broadcom, receiving early access and $100 million in usage credits plus $4 million in open-source security donations to proactively patch vulnerabilities. Separately, CoreWeave announced a cloud infrastructure partnership to help run Anthropic's Claude models. Anthropic also disclosed its annualized run-rate revenue surpassed $30 billion, up from $9 billion at end of 2025, with VCs at HumanX noting sentiment has shifted sharply in its favor versus OpenAI, which is valued at $852 billion compared to Anthropic's $380 billion.

Why it matters

As a product leader and investor in Singapore, Anthropic's tripling of revenue in three months—and the VC sentiment shift at HumanX—signals it is the highest-conviction enterprise AI bet right now; the CoreWeave infrastructure deal and Mythos launch also indicate Anthropic is scaling compute partnerships aggressively, which is relevant when evaluating AI infrastructure plays or enterprise SaaS bets in the region that depend on Claude-class capabilities.

Sources: Business Insider, Techdirt, The Wall Street Journal (first: The Wall Street Journal)

Linux Kernel Adds Official AI Contribution Guidelines Requiring 'Assisted-By' Attribution Tags

The Linux kernel's official documentation now mandates that AI-assisted contributions include an 'Assisted-by' tag with agent name and model version (e.g., 'Assisted-by: Claude:claude-3-opus coccinelle sparse'), while prohibiting AI agents from adding Signed-off-by tags reserved for human DCO certification.

Source: Hacker News

Epoch AI Data Shows U.S. Holds ~75% of Global AI Compute, Google Leads All Entities

Epoch AI's Q4 2025 compute capacity estimates show the U.S. holds roughly 75% of global AI compute with China second at ~10%, and China's total AI compute capacity is roughly equivalent to Oracle's alone, based on H100-equivalent GPU counts.

Source: Business Insider

Semiconductor And Chip Infrastructure

TSMC Reports Record Q1 Revenue of $35.6B, Up 35% Year-on-Year on AI Chip Demand

TSMC reported Q1 2026 revenue of NT$1.13 trillion ($35.6 billion), a 35% year-on-year increase that exceeded analyst forecasts of NT$1.12 trillion. March alone saw a 45.2% year-on-year revenue rise to NT$415.2 billion. SemiAnalysis analyst Sravan Kundojjala said TSMC will 'easily exceed its 30% annual growth target,' attributing the first-quarter beat in part to price hikes on advanced chips and strong AI segment demand that offset weakness in smartphone and PC end markets. TSMC's full Q1 earnings are scheduled for April 16, and ASML reports next week as a semiconductor bellwether. The AI compute concentration data from Epoch AI underscores the demand backdrop: the US holds roughly 75% of global AI compute capacity, with Google's TPUs and Nvidia GPUs dominating.

Why it matters

As a Singapore-based product leader and investor evaluating AI infrastructure plays in APAC, TSMC's sustained 35%+ growth and pricing power on advanced nodes signal continued margin expansion and supply tightness—relevant if you are assessing foundry-dependent fabless chip startups or AI infrastructure investments where TSMC capacity access is a key risk variable. The April 16 full earnings release will clarify gross margin trajectory (SemiAnalysis forecasts 64%) and capacity guidance.

Sources: CNBC, Business Insider (first: CNBC)

SiFive Raises $400M Series G Led by Atreides Management for RISC-V Chip Designs

SiFive, which supplies open RISC-V chip blueprints to customers including Alphabet, raised a $400 million Series G led by Atreides Management—CEO Patrick Little described it as the company's last private round before an IPO.

Source: Crunchbase News

Middle East Conflict Disrupts Helium Supply and Chip Delivery Routes, Damps Regional AI Investment

A month-and-a-half of Iran conflict has curtailed helium exports—a chipmaking input—and caused semiconductor delivery delays to Europe due to flight path disruptions, with analysts warning prolonged fighting could redirect AI data center investment toward APAC, Europe, and Latin America.

Source: CNBC

Ajinomoto Identified as Holding Strategic Monopoly on AI Infrastructure Material

An activist investor has built a stake in Japan's Ajinomoto, arguing the MSG manufacturer holds a lucrative monopoly on a material described as vital to AI chip packaging infrastructure.

Source: The Wall Street Journal

US Tech & Startup News

Anthropic Overtakes OpenAI as Silicon Valley's Favorite at HumanX 2026, Announces $30B Run-Rate Revenue and Mythos Model

At HumanX 2026 in San Francisco's Moscone Center — attended by ~6,700 people paying $4,000+ per ticket — the prevailing sentiment among VCs and founders shifted decisively toward Anthropic and away from OpenAI. Anthropic announced its annualized run-rate revenue surpassed $30 billion, up from $9 billion at end of 2025, representing a roughly 3x increase in three months. The company also unveiled a new frontier model called Mythos, described as too powerful to release publicly due to cybersecurity risks. Anthropic's valuation of $380 billion was cited as more attractive than OpenAI's $852 billion, with Claude Code highlighted as a breakout product. OpenAI faced criticism over its acquisition of internet talk show TBPN and CEO Sam Altman's Pentagon dealings, with one former Coatue/Kleiner Perkins partner predicting a talent brain drain from the company.

Why it matters

As a product leader and investor in Singapore, the Anthropic-OpenAI sentiment shift at HumanX directly affects which platform bets to back in enterprise AI product strategies and portfolio decisions — Anthropic's 3x revenue growth in 90 days and $380B valuation versus OpenAI's $852B changes the risk/return calculus on AI infrastructure vendor choices and investment theses for the region.

Source: Business Insider

SiFive Raises $400M Series G for RISC-V Chip Designs Ahead of IPO

San Mateo-based SiFive raised a $400 million Series G led by Atreides Management for open-standard RISC-V chip designs, with CEO Patrick Little calling it likely the company's final round before an IPO.

Source: Crunchbase News

Amazon CEO Andy Jassy Pledges Project Leo Satellite Internet Launch by Mid-2026 at Up to 1Gbps

Amazon CEO Andy Jassy confirmed in his annual letter that Project Leo satellite internet will enter service by mid-2026, promising download speeds up to 1Gbps — exceeding current Starlink offerings.

Source: Engadget

Google News Surfaces Polymarket Prediction Market Bets Alongside News Stories

Google News began displaying Polymarket betting odds as top results — including in searches about Bitcoin and the Strait of Hormuz — drawing user backlash, leveraging a November partnership with Polymarket and Kalshi that was originally scoped to Google Finance only.

Source: Engadget

Hermeus Raises $200M Equity Plus $150M Debt at $1B Valuation for Autonomous Military Aircraft

El Segundo-based Hermeus raised $350 million total ($200M equity led by Khosla Ventures, $150M debt) at a $1 billion valuation to develop what it claims will be the fastest unmanned defense aircraft.

Source: Crunchbase News

Media, Creator Economy & Adtech

YouTube Premium Raises US Prices Across All Tiers for First Time Since 2023

YouTube has raised subscription prices across all US Premium tiers effective April 10, 2026, for new subscribers and June 2026 for existing ones. The individual Premium plan increases from $13.99 to $15.99/month, the family plan from $22.99 to $26.99/month, Premium Lite from $7.99 to $8.99/month, and YouTube Music from $10.99 to $11.99/month. YouTube cited continued investment in ad-free viewing, background play, and its 300M+ track music library, and noted this is its first price increase since 2023. YouTube reported 125 million combined YouTube Music and Premium subscribers as of March 2025, up from 100 million in 2024. The move follows price hikes by Spotify (February 2026), Netflix (March 2026), and Amazon Prime Video (March 2026).

Why it matters

As a product leader and investor tracking platform monetization in Southeast Asia, this signals that YouTube is willing to extract more subscription revenue from its largest market without adding new content — a strategic bet on lock-in over value expansion. For Singapore-based portfolio companies competing in creator monetization or subscription media, this raises the question of whether regional price increases follow, and whether creator revenue-share mechanics change alongside the subscription uplift.

Sources: Engadget, TechCrunch, CNET, MacRumours, TechRadar, Mashable (first: Engadget)

Onix Launches AI Expert Chatbot Platform Positioned as 'Substack for Bots'

Onix, cofounded by former Wired contributor David Bennahum, launched in beta with 17 vetted health and wellness experts offering AI chatbot versions of themselves at subscription prices, framing expert knowledge as 'a capital asset that generates revenue independent of their time.'

Source: Wired

Alibaba Confirms HappyHorse-1.0 AI Video Model After Anonymous Leaderboard Debut

Alibaba confirmed its ATH AI Innovation Unit built HappyHorse-1.0, which topped Artificial Analysis blind-test rankings for text-to-video and image-to-video generation after a stealth April 7 debut, with Alibaba shares closing up 2.12% on the news.

Source: CNBC