Today in AI

Daily breakdowns of what shipped in AI, written for engineers and the curious. 50 posts.

• 2026-07-13 SK Hynix, the world's second-largest memory maker, warned the AI memory shortage will peak i...

  Today in AI: SK Hynix, the world's second-largest memory maker, warned the AI memory shortage will peak in 2027 and stretch into 2030, on the same day it listed on Nasdaq. This is the constraint most builders keep preten...

• 2026-07-12 Thinking Machines, the lab founded by former OpenAI CTO Mira Murati, published a mission sta...

  Today in AI: Thinking Machines, the lab founded by former OpenAI CTO Mira Murati, published a mission statement arguing frontier labs have the wrong default, that renting one giant model to every customer quietly devalue...

• 2026-07-10 humanoid robots teleoperated by human surgeons removed gallbladders from live pigs, the firs...

  Today in AI: humanoid robots teleoperated by human surgeons removed gallbladders from live pigs, the first end-to-end surgical procedure done by a general-purpose humanoid instead of a dedicated $2M+ surgical robot. The ...

• 2026-07-09 HubSpot now runs 20 billion+ vectors across 200+ indexes and 140+ clusters, serving 38+ inte...

  Today in AI: HubSpot now runs 20 billion+ vectors across 200+ indexes and 140+ clusters, serving 38+ internal teams from a single Vector-as-a-Service platform, and the interesting part isn't the scale, it's how they got ...

• 2026-07-09 Replit shipped a working blueprint for continual learning in production agents, built entire...

  Today in AI: Replit shipped a working blueprint for continual learning in production agents, built entirely at the harness and context layer instead of touching model weights. Most teams run agents on closed frontier mod...

• 2026-07-07 Anthropic published research showing Claude has an internal "J-space", a small, emergent wor...

  Today in AI: Anthropic published research showing Claude has an internal "J-space", a small, emergent workspace where the model does silent multi-step reasoning before it writes a single token. They didn't design it. It ...

• 2026-07-05 Ramp and Revelio Labs pulled data from 21,000 US firms and found the opposite of the layoff...

  Today in AI: Ramp and Revelio Labs pulled data from 21,000 US firms and found the opposite of the layoff narrative. Heavy AI adopters grew headcount 10 percent over two years. Entry-level roles grew 12 percent. The firms...

• 2026-07-03 Apptronik just opened a 90,000 square foot "Robot Park" in Austin where Apollo 2 humanoids r...

  Today in AI: Apptronik just opened a 90,000 square foot "Robot Park" in Austin where Apollo 2 humanoids run logistics, retail, and manufacturing tasks on loop, and the data pipes straight into Google DeepMind. This is th...

• 2026-07-02 Genesis Molecular AI's PEARL model is doing the most interesting diffusion research in the i...

  Today in AI: Genesis Molecular AI's PEARL model is doing the most interesting diffusion research in the industry, and it has nothing to do with LLMs. On the Latent Space podcast, co-founder Evan Feinberg and CTO Sergey E...

• 2026-07-01 Anthropic shipped Claude Sonnet 5 as its new default, priced at $2/$10 per million tokens th...

  Today in AI: Anthropic shipped Claude Sonnet 5 as its new default, priced at $2/$10 per million tokens through August 31, then rising to $3/$15. Same day, the U.S. Department of Commerce lifted export controls on Fable 5...

• 2026-06-30 Meta just shipped Brain2Qwerty v2, a non-invasive brain-to-text model that hit 61 percent wo...

  Today in AI: Meta just shipped Brain2Qwerty v2, a non-invasive brain-to-text model that hit 61 percent word accuracy reading sentences off a magnetoencephalography scan, up from 8 percent on prior non-invasive systems, a...

• 2026-06-29 Zhipu AI just dropped GLM-5.2, a 744B Mixture-of-Experts coding model under MIT license that...

  Today in AI: Zhipu AI just dropped GLM-5.2, a 744B Mixture-of-Experts coding model under MIT license that activates only 40B parameters per token, and Coinbase is already routing production traffic to it. The shift here ...

• 2026-06-28 Coinbase just cut its AI spend nearly in half while token volume kept climbing, and CEO Bria...

  Today in AI: Coinbase just cut its AI spend nearly in half while token volume kept climbing, and CEO Brian Armstrong says the win came from routing, caching, and leaner context, not a smarter model. That's the story this...

• 2026-06-27 Qwen just dropped AgentWorld, a 35B open-weight model that simulates the environments agents...

  Today in AI: Qwen just dropped AgentWorld, a 35B open-weight model that simulates the environments agents act in, terminals, browsers, Android, MCP servers, instead of acting in them. Apache 2.0, on HuggingFace, runs on ...

• 2026-06-26 the White House just told OpenAI to stagger the GPT-5.6 release, gating it to roughly 20 gov...

  Today in AI: the White House just told OpenAI to stagger the GPT-5.6 release, gating it to roughly 20 government-approved partners on Amazon Bedrock before any broader launch. The Information reports the model hit a capa...

• 2026-06-25 OpenAI and Broadcom taped out Jalapeño, OpenAI's first custom inference chip, in nine months...

  Today in AI: OpenAI and Broadcom taped out Jalapeño, OpenAI's first custom inference chip, in nine months, with OpenAI's own models helping design it. This isn't a Nvidia replacement story. It's a vertical integration st...

• 2026-06-24 Anthropic shipped Claude Tag, turning Claude into a tagged Slack teammate that breaks work i...

  Today in AI: Anthropic shipped Claude Tag, turning Claude into a tagged Slack teammate that breaks work into stages, waits on git webhooks for days, and now merges 65% of Anthropic's own product PRs. This is the real shi...

• 2026-06-23 Sakana AI launched Fugu, a model that doesn't generate tokens itself, it routes every reques...

  Today in AI: Sakana AI launched Fugu, a model that doesn't generate tokens itself, it routes every request to a pool of specialist models behind a single OpenAI-compatible API, and claims parity with Anthropic's Mythos a...

• 2026-06-22 John Jumper, the Nobel laureate behind AlphaFold, just left Google DeepMind for Anthropic, d...

  Today in AI: John Jumper, the Nobel laureate behind AlphaFold, just left Google DeepMind for Anthropic, days after Noam Shazeer, co-author of "Attention Is All You Need," defected to OpenAI. Two departures in one week. B...

• 2026-06-21 Cursor, the AI coding IDE built on top of Anthropic and OpenAI models, sold to SpaceX for $6...

  Today in AI: Cursor, the AI coding IDE built on top of Anthropic and OpenAI models, sold to SpaceX for $60 billion this week. Strip away the headline and look at what changed hands. Cursor does not ship a frontier model....

• 2026-06-19 Midjourney, the text-to-image company, just announced a full-body ultrasound scanner.

  Today in AI: Midjourney, the text-to-image company, just announced a full-body ultrasound scanner. No GPU. No model. No AI in the device at all. The Midjourney Scanner lowers you through a ring of 500,000 ultrasound tran...

• 2026-06-15 SpaceX went public at a $1.77 trillion valuation, raising about $75 billion, and the prospec...

  Today in AI: SpaceX went public at a $1.77 trillion valuation, raising about $75 billion, and the prospectus reads like an AI infrastructure thesis with orbital data centers at the core. Musk merged xAI into SpaceX in Fe...

• 2026-06-13 Anthropic pulled Claude Fable 5 and Mythos from every customer worldwide, 3 days after launc...

  Today in AI: Anthropic pulled Claude Fable 5 and Mythos from every customer worldwide, 3 days after launch, following a US government directive flagging a possible jailbreak as a national cybersecurity risk. Anthropic is...

• 2026-06-13 SemiAnalysis maxed out every Anthropic and OpenAI subscription tier and found a $200 per mon...

  Today in AI: SemiAnalysis maxed out every Anthropic and OpenAI subscription tier and found a $200 per month Claude Max user can cost Anthropic up to $8,000 in monthly compute, while ChatGPT Pro can cost OpenAI up to $14,...

• 2026-06-11 Anthropic shipped Claude Fable 5, and the headline is not the benchmark.

  Today in AI: Anthropic shipped Claude Fable 5, and the headline is not the benchmark. It is that Stripe ran a full codebase migration in one day, work that previously took more than two months. 72.9 percent on CursorBenc...

• 2026-06-09 Apple unveiled Siri AI at Tim Cook's final WWDC, and the assistant is partly powered by Goog...

  Today in AI: Apple unveiled Siri AI at Tim Cook's final WWDC, and the assistant is partly powered by Google Gemini models running in Apple's private cloud. This is the rebuild Apple promised two years ago when Apple Inte...

• 2026-06-08 Google just signed a $920 million per month deal with SpaceX to rent 110,000 NVIDIA GPUs thr...

  Today in AI: Google just signed a $920 million per month deal with SpaceX to rent 110,000 NVIDIA GPUs through 2029, roughly $1 billion a month flowing to Elon Musk's rocket company to power Gemini. Let that sink in. Goog...

• 2026-06-07 NVIDIA just dropped the Isaac GR00T Reference Humanoid, an open hardware and software platfo...

  Today in AI: NVIDIA just dropped the Isaac GR00T Reference Humanoid, an open hardware and software platform built with China's Unitree to put humanoid research into university labs. The robot pairs a Unitree H2 body with...

• 2026-06-06 Google just launched Dreambeans, an app that turns your Gmail, Calendar, Photos, YouTube, an...

  Today in AI: Google just launched Dreambeans, an app that turns your Gmail, Calendar, Photos, YouTube, and Search history into a daily set of personalized illustrated stories. Think of it as the opposite of a feed. Inste...

• 2026-06-05 Anthropic just shared that Claude wrote over 80 percent of merged code at the company in May...

  Today in AI: Anthropic just shared that Claude wrote over 80 percent of merged code at the company in May, and their blog post explaining it crossed 3 million views in a day. The report is called "When AI builds itself,"...

• 2026-06-04 Reve and Ideogram both shipped new image models yesterday that quietly change how AI picture...

  Today in AI: Reve and Ideogram both shipped new image models yesterday that quietly change how AI pictures get made. The prompt is no longer the product. Reve 2.0 launched with a code based layout system that lets you fi...

• 2026-06-03 Microsoft just shipped seven in house MAI models at Build 2026, finally writing an AI story...

  Today in AI: Microsoft just shipped seven in house MAI models at Build 2026, finally writing an AI story that isn't OpenAI's. The lineup spans reasoning, code, image, voice, and transcription, led by MAI Thinking 1, MAI ...

• 2026-06-02 Anthropic just confidentially filed its S 1 with the SEC, setting up what could be a near $1...

  Today in AI: Anthropic just confidentially filed its S 1 with the SEC, setting up what could be a near $1 trillion IPO this fall. The Claude maker beat OpenAI to the punch on going public, following a $65B Series H that ...

• 2026-06-01 A startup called MicroAGI is offering free apartment cleanings in New York City, and the cat...

  Today in AI: A startup called MicroAGI is offering free apartment cleanings in New York City, and the catch is not a monthly subscription. It is your data. Here is how it works. A vetted cleaner shows up wearing a head m...

• 2026-05-31 OpenRouter just raised $113M at a $1.3B valuation, and its business model is almost deceptiv...

  Today in AI: OpenRouter just raised $113M at a $1.3B valuation, and its business model is almost deceptively simple. OpenRouter sits between you and over 400 AI models, routes your request to the right one, and takes rou...

• 2026-05-30 Waymo just launched the Ojai, its first purpose built robotaxi, and the engineering choice b...

  Today in AI: Waymo just launched the Ojai, its first purpose built robotaxi, and the engineering choice behind it tells you everything about where autonomous vehicles are heading. Every Waymo you have seen until now star...

• 2026-05-29 Anthropic just shipped Claude Opus 4.8 on the same day it closed a $65B Series H round at a...

  Today in AI: Anthropic just shipped Claude Opus 4.8 on the same day it closed a $65B Series H round at a $965B valuation, making it the most valuable AI lab in the world. The funding number is striking. The model detail ...

• 2026-05-28 Biohub just released ESMFold2, a world model of protein biology, and the implications for me...

  Today in AI: Biohub just released ESMFold2, a world model of protein biology, and the implications for medicine are hard to overstate. Mark Zuckerberg and Priscilla Chan's nonprofit Biohub, led by scientist Alex Rives, d...

• 2026-05-27 A research paper from Sapient Intelligence argues that Chain of Thought reasoning, the techn...

  Today in AI: A research paper from Sapient Intelligence argues that Chain of Thought reasoning, the technique powering most advanced AI today, is architecturally broken. Here is the core idea. Modern LLMs are fixed depth...

• 2026-05-26 A free GitHub tool just bypassed Meta and Google's AI safety filters in under 10 minutes, on...

  Today in AI: A free GitHub tool just bypassed Meta and Google's AI safety filters in under 10 minutes, on a regular laptop. Meta and Google have spent hundreds of millions of dollars building guardrails into their models...

• 2026-05-25 Google DeepMind's AlphaProof Nexus solved nine open Erdős problems autonomously, some untouc...

  Today in AI: Google DeepMind's AlphaProof Nexus solved nine open Erdős problems autonomously, some untouched for 56 years, at a cost of a few hundred dollars each. The architecture is the key story. AlphaProof Nexus comb...

• 2026-05-24 Google's I/O 2026 didn't just ship new models.

  Today in AI: Google's I/O 2026 didn't just ship new models. It shipped a complete vertical stack for the agentic era. Here's what that means. Gemini Omni arrived as the headline, an any-to-any model that handles video, a...

• 2026-05-24 Cursor just hit $3B in annualized revenue, and SpaceX holds a $60B option to acquire it.

  Today in AI: Cursor just hit $3B in annualized revenue, and SpaceX holds a $60B option to acquire it. Three years ago, Cursor didn't exist. Today, over 3,000 enterprises are paying $100,000+ annually for it. That's not v...

• 2026-05-23 Google launched Gemini 3.5 Flash at I/O 2026, and the numbers are hard to ignore.

  Today in AI: Google launched Gemini 3.5 Flash at I/O 2026, and the numbers are hard to ignore. The model runs at four times the speed of competing models while delivering what Google calls frontier-level reasoning. That'...

• 2026-05-21 An OpenAI reasoning model just disproved a math conjecture that stood for 80 years, and it d...

  Today in AI: An OpenAI reasoning model just disproved a math conjecture that stood for 80 years, and it did it for under $1,000. In 1946, Paul Erdős posed a question about unit distances: how many pairs of points in a gr...

• 2026-05-20 Andrej Karpathy just joined Anthropic, and the move tells you everything about where the fro...

  Today in AI: Andrej Karpathy just joined Anthropic, and the move tells you everything about where the frontier is right now. Karpathy is one of the most respected researchers in AI. He cofounded OpenAI, led the Autopilot...

• 2026-05-19 Cursor just shipped Composer 2.5 - a coding model that matches frontier models at a fraction...

  Today in AI: Cursor just shipped Composer 2.5 - a coding model that matches frontier models at a fraction of the cost. Here's what makes this worth paying attention to. Cursor trained Composer 2.5 on 25x more synthetic t...

• 2026-05-18 Anthropic is closing in on a $30B raise at a $900B+ valuation — which would make it the most...

  Today in AI: Anthropic is closing in on a $30B raise at a $900B+ valuation — which would make it the most valuable AI company in the world. Three months ago, Anthropic was valued at $380 billion. As of this raise, co-led...

• 2026-05-17 New federal data shows AI-exposed jobs are shrinking while the rest of the labor market grows.

  Today in AI: New federal data shows AI-exposed jobs are shrinking while the rest of the labor market grows. The Bureau of Labor Statistics released figures today covering roughly 9 million jobs flagged as vulnerable to A...

• 2026-05-14 DeepSeek just dropped V4, and one number stands out: one million tokens of context.

  DeepSeek just dropped V4, and one number stands out: one million tokens of context. To put that in plain terms, most AI models today can hold a conversation about as long as a short novel. DeepSeek V4 can hold one the le...