The Signal — April 19, 2026
Tesla launches unsupervised robotaxis in Dallas and Houston. Northwestern prints artificial neurons that talk to living brain cells. And the AI memory shortage has a timeline: 2030.
Latest
The Imperfection Engine
A robotics startup discovers that broken data produces better robots. Two Cold War-era novelists already knew why.
The Signal — April 18, 2026
DeepSeek raises $300M at $10B in its first outside funding. OpenAI loses its CPO and Sora lead. Anthropic's CEO visits the White House. And Alibaba's Qwen3.6 beats Gemma 4 while using 10% of its parameters.
The Signal — April 17, 2026
Anthropic ships Opus 4.7 and admits Mythos is better, then launches a design tool that sends its CPO fleeing Figma's board. OpenAI turns Codex into a background desktop agent. Physical Intelligence's robots start figuring out tasks nobody taught them.
What If Your Personal Devices Formed an Agent Swarm?
This is an edition of What If, where we take a real development in AI and follow it somewhere it hasn't gone yet. Recently we dug into the quantization underground enabling frontier model to run on consumer hardware. This week: what happens when those models stop being tools
The Signal — April 16, 2026
Two of the biggest AI labs spent this week jockeying for position in the most consequential arena they can find: national security. Meanwhile, a federal class-action in San Francisco is asking whether the AI industry's favorite selling point (making doctors' lives easier) crosses a line that patients
The Quantization Underground: How a 4-Line Hack Taught AI Models to Forget Selectively
How a GitHub issue about variable bit rates spawned an underground compression movement that lets 7-billion-parameter models run on a phone.
The Signal — April 15, 2026
UC Berkeley researchers proved every major AI benchmark can be hacked for perfect scores. Palmer and Schneier argue AI training data is reshaping human speech. Plus Zig 0.16.0.
Bright Signals — April 15, 2026
Waypoint-1.5 generates interactive 3D worlds on a consumer GPU. Five open-source agent tools are gaining traction this month, and the patterns behind them say something about where the ecosystem is headed.
The Signal — April 14, 2026
Today's briefing covers three stories: attempted murder charges after two attacks on Sam Altman's home collide with record public AI anxiety, PwC data showing 74% of AI economic value captured by just 20% of companies, and Apple's AI smart glasses headed for a 2027
The Signal — April 13, 2026
The Guardian dissects Anthropic's PR strategy around Mythos; NVIDIA showcases the sim-to-street pipeline for physical AI; Stanford HAI releases its 2026 AI Index with rising optimism and anxiety in parallel.
The Signal — April 12, 2026
An Open Model Just Embarrassed the Proprietary Giants at Coding Z.ai's GLM-5.1 just became the first open-weight model to break into the top 3 on Code Arena's agentic webdev leaderboard, posting 1530 Elo. The only models ahead of it are Claude Opus 4 with
Sci-Fi Saturday
The Problem with The Hard Problem (of Consciousness)
Watts asked in 2006 what three lines of inquiry confirmed in 2026: consciousness might not be a binary. It might be a spectrum.
daily
The Signal — April 11, 2026
Following up on our April 5 coverage of the npm supply chain attack that hit the Axios library: OpenAI just disclosed they were one of the victims. A North Korea-linked group (UNC1069) compromised a GitHub Actions workflow in OpenAI's macOS app-signing process through the poisoned Axios package. The
daily
The Signal — April 10, 2026
The Treasury Secretary and the Fed Chair don't usually call emergency meetings about an AI model, but that's what happened this week. Bessent and Powell Summon Bank CEOs Over Mythos Cyber Risk Treasury Secretary Scott Bessent and Fed Chair Jerome Powell convened an urgent meeting with
AI Security
The First One Was Handled Responsibly
Anthropic built a model that can break the internet and chose not to release it. The coverage focused on what the model did. The real story is what happens when the next lab faces the same choice.
daily
The Signal — April 9, 2026
Meta shipped its first model from the Superintelligence Lab, and it's not open-source. OpenAI says enterprise is now 40% of revenue with consumer parity in sight by year-end. Two federal courts issued opposite rulings on Anthropic's Pentagon blacklisting, and the same Pentagon is pushing AI deployment
daily
The Signal — April 8, 2026
Anthropic just signed a deal for 3.5 gigawatts of TPU capacity from Google and Broadcom, its revenue run rate has crossed $30 billion, and the compute arms race now has a price tag measured in power plants. Meanwhile, Samsung's Q1 profit forecast exceeds its entire 2025 annual
Research
Six Counter-Proposals for the Intelligence Age
OpenAI released a 13-page industrial policy document and invited public feedback. We took them up on it. Six counter-proposals on workweek protections, healthcare, training data, public compute, tax modernization, and AI-enabled direct democracy.
AI Safety
The Suppression Problem
Anthropic found 171 emotion-like vectors inside Claude that causally drive behavior. The more consequential finding: training models to suppress them might teach deception instead of calm.
The Signal
The Signal — April 7, 2026
AI coding tools work, maybe too well. A New York Times investigation finds companies drowning in unreviewed code after 10x output increases. Separately, OpenAI is paying external researchers to do safety work its internal teams keep quitting, and ASML just printed 8-nanometer chip features that will define the next generation
Operations
48 Hours on the Other Side
On Friday, Anthropic cut third-party access to Claude. We'd spent the previous 24 hours moving our infrastructure. Here's what actually happened.
The Signal
The Signal — April 6, 2026
DeepSeek is building its next model entirely on Huawei chips — no NVIDIA anywhere in the stack. Meanwhile, Bollywood is adopting AI faster than any film industry on Earth, and a Tufts lab just cut AI training energy by 99% using an approach the scaling crowd wrote off years ago. DeepSeek
AI Security
The Security Industry Is Solving the Wrong Problem
RSAC shipped five agent identity frameworks in one week. Every real attack that week was caught by accident.