Search the archive

NEW WAYS TO PLAN TRAVEL WITH AI IN SEARCH

Canvas in AI Mode now builds complete travel itineraries by pulling together flight data, hotel comparisons, Maps reviews, and restaurant recommendations.

READ MORE ON THE METRONOME BLOG

GOOGLE TO RELEASE NANO BANANA PRO POWERED BY GEMINI 3 PRO NEXT WEEK

Google plans to launch Gemini 3 and Nano Banana Pro next week. The company has consistently updated its Gemini models across its suite simultaneously, and the same broad rollout is expected again this time. The Pro branding suggests that Google plans to provide accessible and production-grade generative tools across its platform ecosystem.

WE'RE ROLLING OUT GPT-5.1 AND NEW CUSTOMIZATION FEATURES. ASK US ANYTHING (REDDIT THREAD)

GPT-5.1 is now rolling out to users. It features 8 unique chat styles, making it easier for users to set the tone and style that feels right for them. This Reddit Ask Me Anything features members from OpenAI's team.

OPENAI-BACKED FUTUREHOUSE LAUNCHES KOSMOS AI SCIENTIST

FutureHouse (which is transitioning to commercial spinout Edison Scientific) released an AI scientist that reads 1,500 papers and runs 42,000 lines of analysis code per run. Beta users estimate that a single Kosmos run accomplishes what would take them six months, with 79.4% of conclusions accurate, though the system sometimes pursues statistically significant but scientifically irrelevant findings. It's already made seven discoveries—three reproduced unpublished findings and four contributed novel insights, including one related to Alzheimer's that was validated in human tissue. Kosmos is launching at $200 per run with a generous free tier for academics.

THE 1 BILLION TOKEN CHALLENGE: FINDING THE PERFECT PRE-TRAINING MIX

Most modern language models require massive computational resources and months of training time. This post discusses how to achieve the optimal recipe for creating efficient pre-training data sets to achieve high performance while using dramatically less data. It uncovers critical insights about dataset mixing strategy, failure modes in curriculum testing, and the 'goldilocks zone' for synthetic content. The resulting mixture consistently outperforms complex curriculum strategies and avoids catastrophic failures while maintaining excellent generalization.

THE BITTER LESSONS

A race is a competition with clear boundaries and a clearly defined finish line. This isn't the case with the AI 'race' between the US and China. Nobody currently knows where AI is headed. The field is still being explored. The US and China are taking different strategies that play to their unique strengths. They are locked in a structural conflict, and harmony appears impossible.

that actually match your needs. Get the guide

WE JUST LAUNCHED STRUCTURED OUTPUTS IN THE CLAUDE API

Structured outputs are now available in public beta for Claude Sonnet 4.5 and Opus 4.1. Support for Haiku 4.5 is coming soon. Without structured outputs, Claude may generate malformed JSON responses or invalidate tool inputs, breaking applications, even with careful prompting. Structured outputs constrain Claude's responses so they follow a specific schema. This ensures valid, parsable output for downstream processing.

CONTEXT MANAGEMENT IN AMP

The context window is the entire input that a large language model receives when generating output. It contains messages, the model's replies, tool calls, and the thinking blocks the model outputs to 'reason'. Everything that happens between the user and the model goes into the context window. This article walks users through some of the ways Amp gives users to manage the contents in a context window.

DEPTH ANYTHING 3 RELEASED

Depth Anything 3 (DA3) estimates spatially consistent depth from single or multi-view images, regardless of camera pose availability. It relies on a plain transformer backbone and a single depth-ray representation to simplify architecture and training.

DISTILLATION FOR BLACK-BOX LLMS

GAD introduces a new approach for distilling knowledge from black-box LLMs using only generated text, without access to output probabilities. It trains a student model through adversarial learning against a discriminator, enabling on-policy imitation.

AI IS A NEW COMPUTING PARADIGM

AI allows us to write programs by specifying objectives. If a task is verifiable, then a neural net can be trained to do it extremely well. The more a task is verifiable, the more it is amenable to automation using AI.

SATYA NADELLA'S REFLECTION ON AI PLATFORMS

Microsoft co-designed its new AI superfactory with OpenAI and Nvidia, integrating three generations of frontier training systems and adding AMD for GPT inference to broaden the hardware stack. The company aims to scale this infrastructure so enterprises can train and run their own models without ceding value or control to major AI vendors. The biggest impact will come when industries like pharma, manufacturing, and education use AI to compress timelines and expand output, creating positive-sum economic gains.

HE'S BEEN RIGHT ABOUT AI FOR 40 YEARS. NOW HE THINKS EVERYONE IS WRONG

Yann LeCun may soon be leaving Meta to pursue a startup focused on world models because he believes that large language models are a dead end in the pursuit of superintelligence.

SAKANA AI TAKES CROWN AS JAPAN'S MOST VALUABLE UNICORN

Sakana AI raised $132 million from investors including MUFG, Santander, and US venture firms, pushing its valuation to approximately $2.6 billion, which is the record for an unlisted Japanese startup.

OPENAI SAYS IT'S FIXED CHATGPT'S EM DASH PROBLEM

GOOGLE MAPS RELEASES NEW AI TOOLS THAT LET YOU CREATE INTERACTIVE PROJECTS

Google Maps introduced AI tools, including a builder agent and an MCP server, enabling developers to create interactive projects using Maps data.

AI STARTUP CURSOR RAISES $2.3 BILLION FUNDING ROUND AT $29.3 BILLION VALUATION

Cursor has closed a $2.3 billion funding round at a $29.3 billion post-money valuation. The tool has crossed $1 billion in annualized revenue and swelled to more than 300 employees since it first launched in 2023. Coding tools have proved to be a lucrative AI use case. OpenAI approached Anysphere to potentially purchase Cursor earlier this year, but the deal failed to gain traction.

MURATI'S THINKING MACHINES IN FUNDING TALKS AT $50 BILLION VALUE

Thinking Machines Lab is in early talks to raise new funding at a roughly $50 billion valuation. The new financing would more than quadruple the startup's valuation from July. The deal's terms have yet to be finalized and could change. Thinking Machines aims to build AI models and products to help businesses by supporting more 'human-AI collaboration' across every field of work. The startup has dozens of employees who previously worked at OpenAI.

GOOGLE ADDS DEEP RESEARCH IN NOTEBOOKLM

Google's latest update to NotebookLM introduces Deep Research, a feature designed to support high-quality source integration, and adds support for analyzing new file types, including spreadsheets and Word documents.

NANO BANANA CAN BE PROMPT ENGINEERED FOR EXTREMELY NUANCED AI IMAGE GENERATION

Nano Banana allows for very strong generation control, but it has several issues. It is terrible at style transfer, even though it has essentially no restrictions on intellectual property. The model can be used to create highly specific, high-quality images. This article looks at some of the tricks to using the model.

AI-ORCHESTRATED CYBERATTACK DETAILED BY ANTHROPIC

Anthropic disclosed the first documented AI-driven cyber espionage campaign, carried out largely autonomously using its Claude Code tool. The attackers, linked to a Chinese state-sponsored group, targeted 30+ entities worldwide, demonstrating the growing threat of agentic AI systems in cybersecurity.

GenAI and agentic AI. Read the report

INTRODUCING GPT-5.1 FOR DEVELOPERS

GPT-5.1 uses fewer tokens on simple tasks and shows modest benchmark gains on SWE-bench (+3.5 points to 76.3%) and GPQA Diamond (+2.4 points to 88.1%). On other capabilities, its benchmark scores are flat or slightly lower.

INTRODUCING CODE WIKI: ACCELERATING YOUR CODE UNDERSTANDING

Code Wiki is a platform that maintains a continuously updated, structured wiki for code repositories. It scans the full code base and regenerates documentation after each change, so docs evolve with the code. Every wiki section and chat answer is hyperlinked directly to the relevant code files and definitions. The platform will allow new contributors to start immediately, while senior developers can now understand new libraries in minutes.

WEIGHT-SPARSE TRANSFORMERS HAVE INTERPRETABLE CIRCUITS

OpenAI trained language models with most weights forced to zero, making it easier to identify which neurons link together into a “circuit” representing a certain concept. The approach successfully isolated circuits for tasks like Python quote-matching that are both necessary (removing them breaks the behavior) and sufficient (they work in isolation).

GOOGLE IS ROLLING OUT CONVERSATIONAL SHOPPING—AND ADS—IN AI MODE SEARCH

Google users in the US will soon be able to ask AI Mode complex questions about what to buy. The model will respond with suggestions, guides, tables, and other generated content to help them decide. Some of the content that appears in AI Mode will be ads. There will be an agentic checkout feature that lets AI automatically make purchases with Google Pay at specific retailers. Shopping features are also coming to the Gemini app, but there won't be sponsored content in the results for the time being.

JEVONS OR BUST

Demand for AI is growing, and it seems that it is paying demand. AI is driving incremental revenue to big cloud providers. However, this doesn't mean that AI is definitely a Jevonic-non-bubble instead of a capex binge that ends badly. We have yet to see whether the demand side will materialize.

and efficient LLM running the cloud. See how it works

MEASURING POLITICAL BIAS IN CLAUDE

Anthropic released an open-source "Paired Prompts" evaluation that measures political even-handedness across AI models.

DEEPMIND RELEASED SIMA 2

SIMA 2 is a generalist AI agent that uses Gemini LLM capabilities to reason, adapt, and interact with virtual environments.

AMAZON AND MICROSOFT BACK EFFORT THAT WOULD RESTRICT NVIDIA'S EXPORTS TO CHINA

The Gain AI Act would give tech firms, including Microsoft and Amazon, preferential access to chips at their data centers around the world.

APPLE AND OPENAI MUST FACE XAI LAWSUIT OVER SIRI INTEGRATION

xAI is suing Apple in OpenAI for allegedly conspiring to dominate the AI market.

Test Kimi K2 Thinking today

OPENAI'S GPT-5.1

OpenAI has released its latest models, GPT-5.1 Instant and GPT-5.1 Thinking, which improve performance and enable more customized behavior.

ANTHROPIC'S $50B INFRASTRUCTURE BET

Anthropic is investing $50 billion to build AI data centers in Texas and New York with Fluidstack. This effort supports the US AI Action Plan and reflects a long-term push for domestic AI leadership.

FEI-FEI LI'S WORLD LABS SPEEDS UP THE WORLD MODEL RACE WITH MARBLE, ITS FIRST COMMERCIAL PRODUCT

Marble turns text prompts, photos, videos, or 3D layouts into persistent, editable, and downloadable 3D environments, available now with a free tier. This is the company's first release after raising $230 million and positions the company as the leading “world model” provider for gaming, VFX, VR, and robotics customers.

COMMON GROUND BETWEEN AI 2027 & AI AS NORMAL TECHNOLOGY

Surprisingly, the authors of "AI 2027" (who predict superintelligence within a decade) and "AI as Normal Technology" (who expect gradual diffusion like electricity) agree on many areas. They say most benchmarks will be saturated by 2027-2028 with AI matching expert humans, yet systems will still regularly fail at mundane tasks like "book me a flight to Paris" through 2029 due to long-tail errors. The world will look basically the same in 2029, even with AGI right around the corner.

HERE'S HOW MUCH OPENAI SPENDS ON INFERENCE AND ITS REVENUE SHARE WITH MICROSOFT

OpenAI spent $5.02 billion on inference alone with Microsoft Azure in the first half of 2025. This pattern continued through the end of September. The company's inference costs have risen consistently over the last 18 months. OpenAI's inference costs easily eclipse its revenues.

ALPHAPROOF PAPER

AlphaProof reached silver medal performance in the International Mathematical Olympiad 2024. This post shares details about how the model works. AlphaProof demonstrates how test-time compute can be scaled to obtain improved answers over many orders of magnitude. The researchers developing it were entirely limited by how many TPUs they could get their hands on.