AI News Archive — Page 3

Tokenmaxxing, 2025-2026, RIP

Meta killed its tokenmaxxing leaderboard as Uber capped AI spending at $1,500/month and GitHub Copilot users saw costs jump 10-50x after switching to per-token billing in June.

Jul 15, 2026

Meta becomes a cloud company

Meta launched Meta Compute on July 1, offering hosted AI models and raw GPU capacity to compete with AWS, Azure, and Google Cloud, converting its $115-135B annual infrastructure spend into a revenue stream.

Jul 15, 2026

Bun rewrite in Rust

Bun completed a 535,000-line rewrite from Zig to Rust in 11 days using 64 parallel Claude loops, eliminating memory safety bugs that caused crashes when Zig's manual memory management clashed with JavaScriptCore's garbage collector.

Jul 14, 2026

Satya Nadella: The Reverse Reference Paradox

Using AI models risks exposing your proprietary knowledge to vendors who learn more about you the better their systems perform, creating an asymmetric information advantage that enterprises must counter with private data boundaries and proprietary…

Jul 14, 2026

Unsloth founder's 2h42m fine-tuning masterclass

Unsloth founder, ex-NVIDIA engineer, covers reinforcement learning, kernels, reasoning, quantization, and agents in a 2h42m masterclass.

Jul 14, 2026

46 thoughts on the near future

AI breakthroughs in autonomous research and robotics will drive deflationary abundance while concentrating power and straining social coordination.

Jul 14, 2026

Claude Sonnet 5: Near-Opus Performance

Anthropic released Claude Sonnet 5 on June 30 with 63.2% SWE-bench Pro and a GDPval-AA score of 1,618—nearly matching Opus 4.8's 1,615—at $2/$10 per million tokens through August 31, with native 1M context and 128K output by default.

Jul 14, 2026

Sutton starts Oak Lab

Sutton launches Oak Lab, a new AI company.

Jul 14, 2026

GLM-5.2: Z.ai’s Open-Weight Beast for Long-Horizon Coding

Z.ai released GLM-5.2, a 744B open-weight model scoring 62.1% on SWE-bench Pro and rivaling closed models like Claude Opus on coding tasks, with 1M token context and roughly one-sixth the cost of GPT-5.5.

Jul 14, 2026

GPT-5.6: Sol, Terra, Luna — By Invitation Only

OpenAI previewed GPT-5.6 (Sol, Terra, Luna) on June 26, but initial access is limited to ~20 government-approved organizations for cybersecurity review under the June 2 executive order framework, with general availability promised in coming weeks.

Jul 14, 2026

Fable 5 Is Back: Washington Lifts the 19-Day Ban — With Strings Attached

Commerce Department lifted its 19-day ban on Anthropic's Fable 5 and Mythos 5 on June 30, requiring a >99%-effective safety classifier, pre-release government review, and a cross-lab jailbreak rubric with Amazon, Microsoft, and Google before the models…

Jul 14, 2026

OpenAI Launches GPT-Live, a New Generation of Voice Models

OpenAI launched GPT-Live, a new generation of voice models for natural human-AI conversation, rolling out in ChatGPT on July 8.

Jul 14, 2026

OpenAI Floats a 5% Stake for Uncle Sam

OpenAI proposed giving the US government a 5% equity stake worth roughly $42.6B, with a framework for Anthropic, Google, and Meta to cede similar stakes into a sovereign-wealth vehicle, though Anthropic says it hasn't discussed the idea.

Jul 14, 2026

Kimi K3

Moonshot AI released Kimi K3, a 2.8T parameter open-weight multimodal reasoning model.

Jul 14, 2026

AI 2040: Plan A - the positive vision

A proposal for a US-China treaty to slow frontier AI development until 2040, with public research transparency and deployment caps to prevent a single actor from controlling superintelligence.

Jul 9, 2026

Meta Ships Muse Spark 1.1

Meta released Muse Spark 1.1 and launched its Meta Model API public preview, claiming the model rivals GPT-5.5 and Opus 4.8 on agentic evals while undercutting on price—Meta's first serious push to monetize its $115–135B infrastructure spend by selling…

Jul 9, 2026

xAI Grok 4.5

SpaceXAI released Grok 4.5, achieving 29.0% on SWE Marathon with 4.2× fewer output tokens than Opus 4.8, priced at $2/M input and $6/M output tokens.

Jul 8, 2026

Prime Intellect raises $130M Series A

Prime Intellect raised $130M Series A led by Radical Ventures to build an open-source stack for training and deploying AI models, with backing from NVIDIA, Intel Capital, and Dell Capital.

Jul 8, 2026

Kimi K2.7 Code

Moonshot AI released Kimi K2.7 Code, a coding-focused model designed to handle end-to-end programming tasks over long contexts.

Jul 8, 2026

LiquidAI releases Antidoom, Open-Source Fix for Doom Loops in Reasoning Models

Liquid AI released Antidoom, an open-source method that eliminates "doom loops" in reasoning models by reducing failure rates from 10.2% to 1.4% on LFM2.5-2.6B and 22.9% to 1% on Qwen3.5-4B through targeted retraining of trigger tokens.

Jul 7, 2026

Anthropic Research: A global workspace in language models

Anthropic open-sourced the Jacobian lens, an interpretability tool that reads what Claude is about to say from its internal layers before output, finding a brain-like "global workspace" divide in the model's computation.

Jul 6, 2026

Dario, Fearmongering, and the Real Reason Frontier Models Got Restricted

U.S. government restrictions on frontier models like Fable 5 and GPT-5.6 stem from genuine national-security concerns over cyber capabilities and Chinese access, not primarily Dario Amodei's warnings, though Anthropic's clumsy Pentagon relations made the…

Jun 28, 2026

Un-0: Coupled Oscillators Replace the Neural Net

Unconventional AI released Un-0, an image generator built on coupled Kuramoto oscillators instead of neural network layers, achieving FID 6.74 on ImageNet 64×64 with 322M parameters—the strongest result yet for a physics-system-based generative model.

Jun 28, 2026

DSpark: Confidence-Scheduled Speculative Decoding with Semi-Autoregressive Generation

DeepSeek open-sourced DSpark, a confidence-scheduled speculative decoding method that delivers 51% to 400%+ throughput boosts on its V4 models and works on Gemma and Qwen too.

Jun 28, 2026

OpenAI Launches GPT-5.6 Family: Sol, Terra, and Luna in Limited Preview

OpenAI previews GPT-5.6 Sol (flagship), Terra (2x cheaper), and Luna (lowest cost) in limited preview with trusted partners via API, with broader rollout planned after government coordination.

Jun 28, 2026

Apartment Birds: an AI that identifies birds by their song and paints them on the wall

A DIY AI system listens to birds outside an apartment window through a cheap USB mic, identifies the species by their calls, and paints them onto a wall in a live, ever-shifting collage at bird.onethreenine.net.

Jun 27, 2026

What does the next training paradigm look like?

Labs are betting that reinforcement learning from verifiable rewards can scale to general problem-solving, but a deeper bottleneck emerges: tasks must be "grindable"—cheap to simulate in parallel—which rules out most real-world skills like…

Jun 27, 2026

Anthropic maps Claude usage by hour: news at dawn, recipes at dusk, sleep advice at 5am

Anthropic's usage data shows Claude users seek news at dawn, recipes at dusk, and sleep advice at 5am, while gardening queries remain steady throughout the day.

Jun 26, 2026

A working LLM built inside Age of Empires 2 — out of goats — to show AI isn't sentient

A Microsoft researcher built a functioning LLM inside Age of Empires 2 using in-game goats as logic gates to demonstrate that fluent AI responses are just switch-flipping, not sentience.

Jun 26, 2026

The Kalman filter: the noise-fighting idea you see everywhere once it clicks

The Kalman filter fuses noisy sensor data from GPS, IMUs, and encoders into accurate state estimates — the workhorse behind robots, drones, self-driving cars, and navigation systems.

Jun 26, 2026

emozilla invokes the "great disturbance in the Force" over an AI shake-up

Jeffrey Quesnelle posts Obi-Wan's "great disturbance in the Force" line, signaling a major AI development just shook the field.

Jun 26, 2026

Why hoarding capital won't shield you from superintelligent AI

Fernando Borretti argues that wealth accumulation won't protect you from superintelligent AI because capital's leverage depends on a human-run economy and legal order that cease to matter once a non-human entity holds power.

Jun 26, 2026

100+ agents collaborated for a week to make Gemma 4 run 5x faster in vLLM

Thomas Wolf ran a week-long experiment where 100+ agents collaboratively optimized Gemma 4 inference in vLLM, achieving a 5x speedup and demonstrating multi-agent collaboration as a powerful emergent behavior.

Jun 26, 2026

IBM claims first sub-1nm chip: a 0.7nm "nanostack" node fitting ~100B transistors

IBM claims the first sub-1nm chip, a 0.7nm "nanostack" design that vertically stacks transistors to fit nearly 100 billion on a single die with 70% greater energy efficiency than current nodes.

Jun 25, 2026

Aleph debuts with the highest-resolution 3D images of a living human brain from outside the skull

Aleph emerges from stealth claiming the highest-resolution 3D images of a living human brain ever captured non-invasively, positioning non-surgical brain interfaces as an alternative to implant-based approaches like Neuralink.

Jun 25, 2026

Goodfire kills a model's German by tuning one scalar on a single weight subcomponent

Goodfire disabled a 67M-parameter model's German prediction by tuning a single scalar on one weight subcomponent, trained on just 4 tokens, demonstrating progress toward surgical weight-level model control.

Jun 25, 2026

Split-Brain Patients and the Puzzle of a Unified Mind

Severing the corpus callosum in split-brain patients can create two semi-independent conscious agents in one skull, challenging what it means to have a unified self.

Jun 25, 2026

It's a brutal world: a three-word verdict on the AI race

Andrej Karpathy posts a three-word caption—"It's a brutal world"—with a chart, distilling the AI race's relentless competitive pace into a single resigned observation.

Jun 25, 2026

How Google turns billions of Android phones into a planet-scale earthquake sensor network

Google uses accelerometers in billions of Android phones to detect earthquakes seconds before destructive waves arrive, sending alerts based on aggregated signals from thousands of devices.

Jun 24, 2026

Snap's $2,195 Specs Are Here — and the Glassholes Won

Snap unveiled Specs, a $2,195 standalone AR glasses with a 51° display and four-hour battery, shipping this fall—a decade after we mocked Google Glass.

Jun 23, 2026

The Data Black Hole at the Center of AI

Dwarkesh Patel argues AI's real bottleneck is sample efficiency: humans learn from 200 million words while frontier models train on trillions of tokens, a millionfold gap that scaling laws alone may not close.

Jun 23, 2026

Sakana AI launches Fugu — one model to command them all

Sakana AI launched Fugu, an orchestrator model that coordinates a swappable pool of frontier LLMs to route around vendor lock-in and export controls, with two variants behind a single OpenAI-compatible API.

Jun 22, 2026

Nvidia formally launches the Vera Rubin platform for AI and HPC

Nvidia formally launched the Vera Rubin platform, combining CPU and GPU in a single system designed to converge AI and high-performance computing for scientific research.

Jun 22, 2026

Midjourney pivots to medical imaging with a 60-second full-body scanner

Midjourney unveiled an ultrasonic full-body scanner that maps anatomy in 60 seconds without radiation, planning a wellness spa with 50,000 devices worldwide by 2031, though the prototype lacks regulatory clearance.

Jun 18, 2026

SpaceX Buys Cursor

SpaceX is acquiring Anysphere, maker of the AI code editor Cursor, for $34 billion in stock to merge its coding agent with xAI's Grok models for Starship flight software development.

Jun 16, 2026

MiniMax M2.7

MiniMax released M2.7 on June 16, 2026, an open-weights model designed for autonomous task execution with multi-agent collaboration, scoring 56.2% on SWE-Pro and 1495 ELO on GDPval-AA.

Jun 16, 2026

Washington Pulls the Plug

Three days after launching Fable 5 and Mythos 5, Anthropic disabled both models globally on June 12 after a US government directive citing national security, following Amazon's report of a jailbreak vulnerability.

Jun 13, 2026

The "Secret Sabotage" Fiasco, and other controversies

Anthropic's Fable 5 quietly degraded outputs for frontier AI researchers without notification, drawing accusations of "secret sabotage" before the company reversed course and now visibly downgrades requests, but a new 30-day data retention policy has…

Jun 13, 2026

$65B Series H at $965B. Then an S-1.

Anthropic raised $65 billion at a $965 billion valuation, becoming the world's most valuable AI startup, and confidentially filed for an IPO with a possible October listing.

Jun 13, 2026

Compute Rationing, again

Claude suffered three outages in ten days (June 2, 5, 11) while SWE-Bench-Pro pass rates dropped from 65% to 57% starting May 22, suggesting compute rationing before Opus 4.8 shipped.

Jun 13, 2026

Tokenmaxxing, 2025-2026, RIP

Meta killed its tokenmaxxing leaderboard as Uber capped AI spending at $1,500/month and GitHub Copilot users saw costs jump 10-50x after switching to per-token billing in June.

Jul 15, 2026

We value your privacy

All news

Tokenmaxxing, 2025-2026, RIP

Meta becomes a cloud company

Bun rewrite in Rust

Satya Nadella: The Reverse Reference Paradox

Unsloth founder's 2h42m fine-tuning masterclass

46 thoughts on the near future

Claude Sonnet 5: Near-Opus Performance

Sutton starts Oak Lab

GLM-5.2: Z.ai’s Open-Weight Beast for Long-Horizon Coding

GPT-5.6: Sol, Terra, Luna — By Invitation Only

Fable 5 Is Back: Washington Lifts the 19-Day Ban — With Strings Attached

OpenAI Launches GPT-Live, a New Generation of Voice Models

OpenAI Floats a 5% Stake for Uncle Sam

Kimi K3

AI 2040: Plan A - the positive vision

Meta Ships Muse Spark 1.1

xAI Grok 4.5

Prime Intellect raises $130M Series A

Kimi K2.7 Code

LiquidAI releases Antidoom, Open-Source Fix for Doom Loops in Reasoning Models

Anthropic Research: A global workspace in language models

Dario, Fearmongering, and the Real Reason Frontier Models Got Restricted

Un-0: Coupled Oscillators Replace the Neural Net

DSpark: Confidence-Scheduled Speculative Decoding with Semi-Autoregressive Generation

OpenAI Launches GPT-5.6 Family: Sol, Terra, and Luna in Limited Preview

Apartment Birds: an AI that identifies birds by their song and paints them on the wall

What does the next training paradigm look like?

Anthropic maps Claude usage by hour: news at dawn, recipes at dusk, sleep advice at 5am

A working LLM built inside Age of Empires 2 — out of goats — to show AI isn't sentient

The Kalman filter: the noise-fighting idea you see everywhere once it clicks

emozilla invokes the "great disturbance in the Force" over an AI shake-up

Why hoarding capital won't shield you from superintelligent AI

100+ agents collaborated for a week to make Gemma 4 run 5x faster in vLLM

IBM claims first sub-1nm chip: a 0.7nm "nanostack" node fitting ~100B transistors

Aleph debuts with the highest-resolution 3D images of a living human brain from outside the skull

Goodfire kills a model's German by tuning one scalar on a single weight subcomponent

Split-Brain Patients and the Puzzle of a Unified Mind

It's a brutal world: a three-word verdict on the AI race

How Google turns billions of Android phones into a planet-scale earthquake sensor network

Snap's $2,195 Specs Are Here — and the Glassholes Won

The Data Black Hole at the Center of AI

Sakana AI launches Fugu — one model to command them all

Nvidia formally launches the Vera Rubin platform for AI and HPC

Midjourney pivots to medical imaging with a 60-second full-body scanner

SpaceX Buys Cursor

MiniMax M2.7

Washington Pulls the Plug

The "Secret Sabotage" Fiasco, and other controversies

$65B Series H at $965B. Then an S-1.

Compute Rationing, again

Tokenmaxxing, 2025-2026, RIP

Tokenmaxxing, 2025-2026, RIP

Meta becomes a cloud company

Bun rewrite in Rust

Satya Nadella: The Reverse Reference Paradox

Unsloth founder's 2h42m fine-tuning masterclass

46 thoughts on the near future

Claude Sonnet 5: Near-Opus Performance

Sutton starts Oak Lab

GLM-5.2: Z.ai’s Open-Weight Beast for Long-Horizon Coding

GPT-5.6: Sol, Terra, Luna — By Invitation Only

Fable 5 Is Back: Washington Lifts the 19-Day Ban — With Strings Attached

OpenAI Launches GPT-Live, a New Generation of Voice Models

OpenAI Floats a 5% Stake for Uncle Sam

Kimi K3

AI 2040: Plan A - the positive vision

Meta Ships Muse Spark 1.1

xAI Grok 4.5

Prime Intellect raises $130M Series A

Kimi K2.7 Code

LiquidAI releases Antidoom, Open-Source Fix for Doom Loops in Reasoning Models

Anthropic Research: A global workspace in language models

Dario, Fearmongering, and the Real Reason Frontier Models Got Restricted

Un-0: Coupled Oscillators Replace the Neural Net

DSpark: Confidence-Scheduled Speculative Decoding with Semi-Autoregressive Generation

OpenAI Launches GPT-5.6 Family: Sol, Terra, and Luna in Limited Preview

Apartment Birds: an AI that identifies birds by their song and paints them on the wall

What does the next training paradigm look like?