SIGNAL by Prompeteer.ai
Anthropic built an AI model too dangerous to release. Intel joined Musk's $25B Terafab. A training-data giant got supply-chain attacked. Robot taxes proposed. ServiceNow went all-in on agents. Perplexity's pivot drove 50% ARR growth in a month. The EU quietly rolled back AI protections. Meta locked in $21B for NVIDIA Rubin capacity. Eight stories — plus funding rounds, radar signals, and a fake disease that fooled every major LLM.
Anthropic Built an AI That Finds Zero-Days in Every Major OS. It's Too Dangerous to Release.
Claude Mythos Preview, Anthropic's unreleased frontier model, has identified thousands of previously unknown zero-day vulnerabilities in every major operating system, every major web browser, and a range of critical software — including a 27-year-old bug in OpenBSD and a 16-year-old flaw in FFmpeg that survived five million automated test runs. The model wrote a browser exploit chaining four vulnerabilities, escaping both renderer and OS sandboxes. It converts 72.4% of discovered vulnerabilities into working exploits in Firefox's JavaScript shell. Anthropic won't make Mythos generally available. Instead, it launched Project Glasswing — a restricted partnership with Amazon, Apple, Google, Microsoft, NVIDIA, and Palo Alto Networks to secure critical infrastructure before models with similar capabilities proliferate.
Our Take: First credible demo that AI capability has crossed a security threshold. Anthropic made the right call restricting access, but the window is narrow — other labs will reach this within months.
Intel Joins Musk's $25B Terafab Megaproject. Foundry Dominance at Stake.
Intel signed on as foundry partner for Terafab, the joint venture between Tesla and SpaceX. The partnership claims it will deliver one terawatt of annual AI compute from an Austin megafab. Intel is contributing its 18A process node, a 1.8nm technology that's the most advanced U.S.-made semiconductor process. Total project cost: $20–25 billion. Analysts remain skeptical the budget will match the ambition, but the deal represents the biggest external customer win for Intel CEO Lip-Bu Tan. It's a direct challenge to NVIDIA's compute dominance and a bet-the-company moment for Intel's foundry business.
Our Take: One-terawatt target is aspirational. But Intel's 18A is real, and Tan's desperation to prove the foundry model works makes this credible. If it delivers even 20% of capacity, it reshapes the U.S.-China compute balance.
AI Training Data Goldmine Breached Via Open-Source Dependency. The First Real Supply Chain Attack.
Mercor, a $10 billion startup that provides AI training data to the world's top frontier labs, confirmed a devastating supply chain attack through LiteLLM, an open-source library downloaded millions of times daily. The hacking group TeamPCP planted malicious code by compromising the maintainer's credentials. One major customer immediately froze its partnership. 4 terabytes of data were reportedly stolen, including database records, source code, and customer records. The Lapsus$ group claimed involvement. This is the attack the industry was warned about — AI companies built their competitive moats on training data, and a single compromised open-source dependency just exposed the entire supply chain.
Our Take: This is the attack the industry was warned about. AI companies built their moats on training data — and a compromised open-source dependency just exposed the entire supply chain. Every company using LiteLLM should be running forensics right now.
OpenAI Proposes Robot Taxes, Wealth Funds, Four-Day Week. Economic Blueprint Released Pre-IPO.
OpenAI published a 13-page policy paper titled "Industrial Policy for the Intelligence Age" with 11 substantive proposals: taxes on automated labor, a nationally managed public wealth fund seeded by AI companies, formal worker voice in AI deployments, and subsidies for power grid infrastructure. The timing is strategic — released at an $852 billion valuation right before the company's IPO roadshow. The pull quote that defines the moment: "We are entering a new phase of economic and social organization that will fundamentally reshape work, knowledge, and production."
"We are entering a new phase of economic and social organization that will fundamentally reshape work, knowledge, and production." — OpenAI
Our Take: Substantive proposals, strategic timing. OpenAI framing itself as responsible steward of superintelligence right before asking public markets for a trillion-dollar valuation. The paper reads better as an S-1 preamble than a policy brief.
Perplexity Pivots to Agents, ARR Jumps 50% in One Month. The Search-to-Agents Shift Works.
Perplexity's annual recurring revenue topped $450 million in March, a 50% jump in a single month and double quarter-over-quarter growth. The company shipped Computer for desktop automation and Model Council for multi-model comparison. It shifted to usage-based pricing with credits to handle agent workloads. With 100 million monthly active users and an ~$20 billion valuation, internal targets put ARR at $656 million by year-end. The real signal: usage-based pricing is the future of agentic AI — agents consume unpredictable compute, flat subscriptions can't capture that value.
Our Take: Perplexity just proved the search-to-agents pivot in real revenue. Usage-based pricing is the tell — agents consume unpredictable compute, flat subscriptions can't capture that value. Every AI company still on per-seat pricing should be watching.
ServiceNow Declares Every Product AI-Native. Context Engine Becomes Core.
ServiceNow announced a sweeping overhaul: every service, platform, and product in its portfolio is now AI-enabled with agentic automation. The centerpiece is the Context Engine, which pairs agents with institutional knowledge bases. Supporting components include the AI Control Tower and Workflow Data Fabric. A unified EmployeeWorks interface surfaces it all. Most significantly, ServiceNow is opening its SDK to Claude Code, Codex, and Cursor. The company also launched Enterprise Service Management Foundation to serve SMBs. The real story: Context Engine. Agents without institutional memory are chatbots with API access. Opening the SDK to best-in-class dev tools means letting them build on your platform and becoming the default substrate.
Our Take: Context Engine is the real story. Agents without institutional memory are chatbots with API access. Opening the SDK to Claude Code and Codex is smart distribution — let the best dev tools build on your platform, become the default substrate.
EU Proposes Rolling Back AI Rights Protections. Before Enforcement Even Begins.
Amnesty International warned that the EU's "simplification" of the AI Act and GDPR will weaken core safeguards. 127 civil society organizations signed an open letter opposing it. The Commission is redefining personal data, relaxing consent requirements, and postponing key AI Act elements by a year. The simplification package narrows the personal data definition, potentially letting Big Tech harvest more data for AI training. The EU AI Act's full enforcement is due in August 2026 — just four months away. Critics call it "covert dismantling of Europe's strongest digital protections."
Our Take: The EU spent years building the world's most comprehensive AI regulatory framework, and is now weakening it before enforcement begins — under pressure from the same companies it was designed to regulate. If the simplification package passes as drafted, the AI Act becomes a framework with escape hatches large enough to drive a data center through.
Meta Locks in $21B for NVIDIA Rubin Capacity. Leasing the Agentic Future.
Meta struck a $21 billion deal to secure early-access capacity on NVIDIA's next-generation Rubin GPU clusters through CoreWeave, bypassing its own data center timelines to accelerate deployment of Llama 5 agentic models expected in late 2026. The deal represents a critical shift: rather than waiting years for in-house infrastructure, Meta is leasing cutting-edge compute to stay in the agentic race now. CoreWeave — valued at $35 billion after its March IPO — is becoming the compute broker of the AI era. The strategic read: Meta lost the foundation model lead to closed-source competitors and is spending its way back in through infrastructure speed, not model elegance.
Our Take: Build vs. lease used to be an infrastructure question. Now it's a competitive strategy. Meta is renting its way to Llama 5 because waiting to build means losing the agentic window entirely. CoreWeave just became the AWS of the GPU era.
Rapid Fire
All 11 original cofounders of Elon Musk's xAI have now departed, with the last two — Manuel Kroiss and Ross Nordeen — leaving in March. Musk said xAI "was not built right first time around, so it is being rebuilt from the foundations up." The fastest complete founding-team exodus in big tech AI history.
Eli Lilly — 1,016 Blackwell Ultras
Eli Lilly launched LillyPod, the world's first NVIDIA DGX SuperPOD with DGX B300 systems: 1,016 Blackwell Ultra GPUs delivering 9,000+ petaflops for drug discovery. Assembled in four months. Workloads span genomics, molecule design, clinical trials, and manufacturing — pharma's compute arms race just entered a new tier.
State Legislatures — 78 Chatbot Bans
78 laws across 27 states now ban AI therapy bots. Maine sent its ban to the governor this week, Missouri's House approved a $10,000-penalty version, Tennessee signed its into law. The Future of Privacy Forum is tracking 98 chatbot bills across 34 states. Regulation is outpacing deployment.
Jack Dorsey's Block eliminated 4,000 jobs — 40% of its global workforce — in one of Q1's largest tech layoffs. The company is restructuring around AI-first operations. Combined with Oracle's 30K cuts last month, big tech is quietly replacing headcount with compute.
On Our Radar
AI Chatbots/Apps: ChatGPT Launches on Apple CarPlay
ChatGPT Voice launched on Apple CarPlay on April 2, the first major AI assistant in a new CarPlay app category for "voice-based conversational apps." It's voice-only, sandboxed, and can't control any vehicle or phone functions. Apple quietly opened the slot — competing assistants like Gemini and Claude can ship into it whenever ready.
Agents: Gartner Projects $201.9B Spend in 2026
Gartner forecasts agentic AI spending will reach $201.9 billion in 2026, a 141% increase year-over-year. By 2027, it overtakes chatbot and assistant spending for the first time. The 119% CAGR to $752.7 billion by 2029 tells the story — agents aren't a feature, they're the next platform.
Workflows: Google Ships Gemma 4 Under Apache 2.0
Google released the Gemma 4 model family on April 2 — lightweight, open, optimized for agents, and runnable on a single GPU or laptop. Under Apache 2.0 license, it's free for commercial use. Designed as the edge counterpart to Gemini's cloud, enabling hybrid AI architectures that are private, low-latency, and cost-efficient. Open-source AI just got meaningfully more accessible.
Legal: Federal Court Rules Public AI Tools Strip Attorney-Client Privilege
In United States v. Heppner, a federal judge ruled that documents generated using consumer-grade AI tools are not protected by attorney-client privilege or work product doctrine. The logic: publicly available AI platforms compromise confidentiality by design. Law firms now face a hard fork — enterprise legal AI with auditability and data isolation, or malpractice exposure. Every general counsel should be rewriting their AI governance policy this week.
Marketing: CMO Survey Shows AI Usage Doubled — But Budgets Are Shrinking
The 2026 CMO Survey reveals AI usage in marketing nearly doubled in two years — from 13.1% to 24.2% — while generative AI grew 220%. The catch: marketing budgets fell to 9.0% of revenue and training spend dropped to 3.8%. Only 6.3% of companies have AI agents in full production. The gap between adoption enthusiasm and operational readiness is the defining tension for marketing leaders this year.
Fundings
Eclipse's new fund backs and incubates startups blending AI with robotics, autonomous systems, and manufacturing hardware — split between a $591M early-stage incubation fund and a growth-stage vehicle. The bet: the next wave of AI value lives in the physical world.
Hermeus — $350M Series, $1B Valuation
The LA-based startup building AI-autonomous hypersonic aircraft hit a $1B valuation. Khosla Ventures led. Founders Fund and In-Q-Tel also participated — defense AI is pulling serious venture capital.
Xoople (Spain) — $130M Series B
Building a satellite constellation to collect geospatial data for AI models, with a deal with L3Harris to build sensors. Led by Nazca Capital. Europe's AI infrastructure play is real — and it's pointed at Earth, not the cloud.
Zero Shot — $100M Fund, Frontier Lab Alumni
Five former frontier-lab staffers (Evan Morikawa, Andrew Mayne, Shawn Jain, Kelly Kovacs, Brett Rounsaville) launched an early-stage AI fund. The operator-to-investor pipeline that took a decade in SaaS takes two years in AI.
Nexthop AI — $500M Series, AI Infrastructure
AI infrastructure developer raised one of Q1's largest non-frontier-lab rounds, signaling the picks-and-shovels layer of the AI stack is attracting serious capital beyond model builders.
Believe It or Not
Scientists at the University of Gothenburg invented a fake disease called "bixonimania" — described as eyelid discoloration from blue light exposure — and uploaded two obviously bogus papers to a preprint server. The papers referenced Starfleet Academy and explicitly stated the work was "made up." Within weeks, ChatGPT, Gemini, and Copilot were diagnosing it as real and advising users to see an ophthalmologist. The misinformation then leaked into actual published medical literature: a paper in Cureus cited bixonimania until Nature flagged it and the journal retracted. Two fake preprints. Zero review. Global medical hallucination.
The Number That Matters
$300 billion — Global venture investment in Q1 2026 alone — more than any full year in history. AI claimed 80% of it. Four companies took 65%. This isn't a rising tide lifting all boats. It's a capital supernova concentrated in a handful of frontier labs, and the gravitational pull is warping the entire venture ecosystem around a single thesis: that whoever builds the most capable model wins everything. The concentration is the story: startups outside frontier AI are finding the funding well increasingly dry, even as headline totals shatter records.
The Social Scene
X — Deepfakes Flood Iran-US Coverage
AI-generated deepfake images and videos about Iran-US tensions flooded X this week despite the platform's content policy crackdown. The volume overwhelmed moderation — a preview of what every election cycle will look like from now on.
LinkedIn — Now #2 AI Search Source
LinkedIn now ranks #2 in citations across major AI search platforms, appearing in 11% of AI responses. Brands are scrambling to optimize their presence for chatbot retrieval, not human feeds — a new SEO discipline is being born.
HackerNews — FDA's AI "Breakthrough" Standard Rising
The FDA's evolving stance on AI "breakthrough" medical devices sparked front-page debate. The agency is now prioritizing AI that solves problems physicians can't — not just tools that improve existing workflows. The standard for "breakthrough" is rising fast.
Tweet of the Week
Elon Musk
@elonmusk
Grok can help you come up with great prompts for images and videos
The post triggered thousands of user-generated examples flooding X, turning into a spontaneous demo reel for Grok Imagine. What made it work wasn't the technology — it's the distribution. When the CEO IS the growth channel with 200M+ followers, traditional marketing budgets become irrelevant.
The Hire / The Fire
HIRED: Stephanie Goutos — First Chief AI Officer at Littler
Littler, the world's largest employment and labor law firm, appointed its first-ever Chief AI Officer. Goutos will oversee AI strategy, governance, and responsible AI deployment across all practice areas globally. When BigLaw creates a C-suite AI seat, the profession's transformation isn't theoretical anymore.
HIRED: David Rice — First Chief AI Officer at HSBC
HSBC, one of the world's largest banking institutions, created and filled its first Chief AI Officer role effective April 1. Rice will lead enterprise-wide AI adoption across global banking operations. Finance is moving AI from innovation labs to the org chart.
HIRED: Anuj Kadyan — McKinsey to CBRE as CTO
McKinsey senior partner joins CBRE Group as Chief Technology & Transformation Officer, effective May 15. One of the world's largest real estate services firms is betting its digital future on AI-native operations. Consulting-to-operator pipeline is accelerating.
FIRED: Block Eliminates 4,000 Jobs — 40% of Workforce
Jack Dorsey's Block cut 40% of its global headcount in one of Q1 2026's most dramatic restructurings. The company is rebuilding around AI-first operations. Q1 2026 tech layoffs now total 80,000 — half attributed directly to AI replacement.
The Reality Check
A February 2026 NBER study surveyed thousands of firms and found 90% report zero measurable impact of AI on workplace productivity — while those same executives projected AI would boost productivity by 1.4% and output by 0.8%. The rhetoric-to-results gap is the most underreported story in enterprise AI. Companies are deploying agents, buying seats, and restructuring organizations around a technology whose measured productivity impact, for the vast majority of adopters, remains statistically indistinguishable from zero. This doesn't mean AI won't deliver — but it means the timeline is longer than the investment thesis assumes.
Until next Friday —
The SIGNAL Team
Prompeteer.ai
© 2026 GARAGE CAPITAL LLC. ALL RIGHTS RESERVED.