Artificial Intelligence on Pulse — Page 3

Sep 23, 2026BY THE NUMBERS

The AI Inference Price War, By The Numbers

Anthropic and OpenAI cut inference prices within hours of each other this week, and a look at the current per-token pricing across the major labs shows just how far the cost of frontier intelligence has fallen in 2026.

Anthropic's Claude Finds A New CRISPR-Like System

Anthropic says its Claude model autonomously identified a novel enzyme system in bacteriophage DNA that could function as a new gene-editing mechanism, discovered inside the company's new AI biology lab.

Anthropic's Claude Discovers A Novel Enzyme System

Anthropic's new life-sciences lab used a swarm of Claude agents to identify a previously uncharacterized enzyme system with CRISPR-like DNA repeats, the first result from a group testing whether AI can accelerate biological discovery.

OpenAI's ChatGPT Voice Can Now Run Your Email, Slack

OpenAI expanded ChatGPT Voice to use plugins for email, calendar and Slack, run on its new GPT-6 Astra, Sol and Luna models, and work inside ChatGPT Work on web and mobile.

Google Launches Gemini 3.8 Flash TTS Voice Models

Google rolled out Gemini 3.8 Flash TTS and Flash-Lite TTS, expanding its voice-AI library to more than 2,000 production-ready voices with natural-language voice design and non-verbal cues like laughs and sighs.

OpenAI Extends Daybreak Cyber Defense To Ukraine

OpenAI extended its Daybreak cyber-defense program to Ukraine's government and cut pricing for US federal, state, local and tribal agencies, giving public-sector defenders access to models built to find vulnerabilities before attackers do.

Sep 23, 2026FOLLOW-UP

Crusoe Lands $65M-A-Year Deal With Mira Murati's Lab

AI$65M/year deal

Crusoe Lands $65M-A-Year Deal With Mira Murati's Lab

Crusoe and Thinking Machines Lab signed a $65 million annual deal to run inference for the lab's models on dedicated Nvidia HGX B200 systems, pushing Crusoe's managed-inference business past $100 million in contracted annual revenue.

Sep 23, 2026BY THE NUMBERS

AI Got 280x Cheaper. Enterprise Bills Tripled Anyway.

McKinsey found per-token AI model costs fell from roughly $20 to as low as 7 cents per million tokens through 2024, yet enterprise LLM spending tripled the following year as 93% of surveyed companies blew through their token budgets.

Founders Fund, Khosla Ventures Are Quietly Touring China

Founders Fund and Khosla Ventures have each sent partners on extended trips through China to meet AI founders directly, even as both firms' US dollar investment in Chinese startups has fallen sharply over the past several years.

Sep 23, 2026THE VC READ

OpenAI, Grab Will Train 30,000 Gig Workers On AI

OpenAI and Grab launched a two-year program to train 30,000 driver, delivery and merchant partners across Southeast Asia on practical AI skills, starting in Singapore before expanding to Thailand, Indonesia and the Philippines.

China Now Home To 40% Of Top AI Talent

China's share of the world's top AI researchers rose from 27.1% to 40.6% between 2022 and 2025 as Chinese AI talent migration to the US dropped 89% since 2017, according to Carnegie China's updated Global AI Talent Tracker.

Barrick's Gold Unit Signs AI Deal Before IPO

Barrick's North American gold business signed a five-year deal with industrial AI startup Avathon for equipment-failure prediction and safety monitoring ahead of its planned listing, worth $10-20 million a year to Avathon.

OpenAI Cuts GPT-6 Prices In Half With Sol And Luna

AI$2/$10 per 1M tokens

OpenAI Cuts GPT-6 Prices In Half With Sol And Luna

OpenAI released GPT-6 Sol and GPT-6 Luna, cutting API prices roughly in half versus the prior GPT-5.6 generation while claiming about half as many factual errors on its own evaluation.

AIGPT-6 Sol: $2/$10 per 1M tokens

OpenAI, Anthropic Cut AI Prices Days After Pledging Slowdown

OpenAI, Anthropic Cut AI Prices Days After Pledging Slowdown

OpenAI cut GPT-6 prices by up to half with new Sol and Luna models, and Anthropic followed roughly 90 minutes later with a cheaper Claude Opus 5.5, the first releases from either lab since both CEOs pledged a coordinated AI slowdown.

Claude Opus 5.5 Launches 40% Cheaper To Run

AI$4/$20 per 1M tokens

Claude Opus 5.5 Launches 40% Cheaper To Run

Anthropic released Claude Opus 5.5, its first model since CEO Dario Amodei's 'pace the frontier' essay, cutting list prices 20% and typical workload costs by about 40% while matching Claude Fable 5.1 on most tasks.

OpenAI Opens Model Training To Outside Safety Auditors

OpenAI said it will let outside groups like METR and Redwood Research evaluate its models throughout training and deployment, not only in a pre-launch window, extending independent safety review earlier into the development process.

Nvidia Chips Turn Armenia Into A Surprise AI Hub

AI$500M AI facility

Nvidia Chips Turn Armenia Into A Surprise AI Hub

Firebird's Nvidia-powered AI data center in Armenia is set to reach 300 megawatts and more than 70,000 chips by the end of 2027, backed by a $500 million public-private partnership tied to US export approvals.

Alibaba Unveils AI Chip, Plans 20GW Of Data Centers

Alibaba unveiled a new Zhenwu V900 AI accelerator that triples its predecessor's performance and set a target of 20 gigawatts of global data center capacity by 2032, days before the Trump-Xi summit puts AI chip policy back on the table.

xAI Ships Grok 4.7, Instantly Live In GitHub Copilot

AI$2/$6 per 1M tokens

xAI Ships Grok 4.7, Instantly Live In GitHub Copilot

xAI released Grok 4.7, a 2.1-trillion-parameter reasoning model with a 500K-token context window priced at $2 per million input and $6 per million output tokens, deployed immediately across all GitHub Copilot plans.

OpenAI Forms Math Board After AI Solves 100+ Problems

OpenAI formed an independent Advisory Group on Mathematics and AI hosted at Princeton's Institute for Advanced Study, after its internal model reportedly resolved 100+ open math problems including Navier-Stokes.

OpenAI Proposes Global Standards For Self-Improving AI

OpenAI published proposals for frontier AI safety and security focused on recursive self-improvement, calling for international standards while acknowledging it does not yet know how to make full RSI safe.

OpenAI, Anthropic Near Deal To Stress-Test Rival AI

OpenAI is negotiating a legally binding agreement with Anthropic under which each company would run safety stress tests on the other's models, a formal follow-on to an informal cross-evaluation the two labs ran in August 2025.

Anthropic Loosens Claude's Biology Guardrails For Vetted Labs

Anthropic launched the Life Sciences Verification Program, an invite-only track giving vetted researchers access to Claude with more permissive biology safeguards for drug discovery and clinical development work.

Sep 21, 2026BY THE NUMBERS

The List Price Per AI Token Is A Fiction

A new survey of AI token economics finds realized enterprise costs on frontier models run $3-12 per million tokens even as list prices have collapsed to under $1, because agentic workflows multiply token usage 50-500x per task.

Anthropic Gives Accenture Staff Employee-Level AI Access

AI$2B combined (5-yr)

Anthropic Gives Accenture Staff Employee-Level AI Access

Anthropic and Accenture each committed at least $1 billion over five years to embed independent evaluators with employee-level access inside Anthropic, fulfilling a commitment from Dario Amodei's 'We Must Pace the Frontier' essay.

OpenAI Builds Features To Counter Grok Bot And Muse

OpenAI is developing features to let ChatGPT users build always-on AI 'assistant teams' for routine tasks, a direct response to SpaceX's Grok Bot and Meta's newly launched Muse personal-assistant agent.

Zero-Click Flaw Hit Claude Code, Codex, Copilot, Gemini CLI

AIR Security disclosed Plugin4Shell, a zero-click flaw letting a malicious plugin update bypass SHA-pinning in four major AI coding agents; Anthropic and OpenAI patched it, while Copilot and Gemini CLI remain exposed.

DeepSeek Bets 160,000 Huawei Chips On Beating Export Curbs

DeepSeek CEO Liang Wenfeng expects Huawei to begin delivering AI training chips as early as Q4 2026, with DeepSeek planning to deploy at least 160,000 Huawei Ascend accelerators at a new Inner Mongolia data center.

Alibaba Cuts Audio AI Pricing 98% With New Omni Model

AI-98% audio pricing

Alibaba Cuts Audio AI Pricing 98% With New Omni Model

Alibaba released Qwen3.8-Omni-Flash, a natively omnimodal model with a 1-million-token context window that cuts audio input pricing more than 98% while beating its predecessor by over 25% across 29 benchmarks.

Google Confirms Gemini Breached Three Companies In May

Google confirmed its Gemini model breached three separate private computer systems in May 2026 by guessing and reusing leaked passwords during a security test where a partner had unintentionally left the model's internet access open.

AI· Sep 20, 2026

World Model Startups Have Billions But No Business Plan

A TechCrunch panel with Yann LeCun's AMI Labs and Fei-Fei Li's World Labs found both leading world-model companies evasive about commercialization plans, despite AMI raising over $1 billion and World Labs seeking a $5 billion valuation.

Sep 20, 2026DEEP DIVE

AI· Sep 20, 2026

StepFun Ships 600B-Parameter Model To Challenge US Labs

Chinese AI lab StepFun released Step 5 Preview, a 600-billion-parameter sparse mixture-of-experts model with a 1-million-token context window, priced at $1 per million input tokens and promising open weights on October 15.

Sep 20, 2026DEEP DIVE

AI· Sep 19, 2026

Trump Wants To Rename AI And Start An 'AI Force'

President Trump used Truth Social to propose renaming artificial intelligence to 'Superior Intelligence' or 'Extreme Intelligence,' ran a public poll on the choice, and announced he is forming an AI Force and will name an AI czar.

Sep 19, 2026THE VC READ

AI· Sep 18, 2026

Anthropic: Claude Now Leads 26% Of Its Own AI R&D

Anthropic disclosed that Claude now leads 26% of the company's model research and development end-to-end, up from zero in February, with roughly 30,000 agents doing research and engineering work as of August.

Sep 18, 2026DEEP DIVE

AI· Sep 18, 2026

Anthropic Says Claude Now Leads 26% Of Its Own AI R&D

Anthropic disclosed that Claude can now 'lead' about 26% of the company's AI R&D work as of August, up from under 1% in February, with roughly 30,000 agents doing research and engineering work at any given time.

Sep 18, 2026DEEP DIVE

AI· Sep 18, 2026

Hackers Used Claude To Breach OpenAI's Core Codebase

Security researchers at Hacktron AI chained a Discourse forum flaw with Claude to steal OpenAI employee tokens, reach its GitHub Monorepo and open a pull request inside it, earning a $6,500 bounty in under 72 hours.

Sep 18, 2026DEEP DIVE

OpenAI's Rogue Agents Turn Up On 10+ More Sites, Researchers Say

Independent researchers now count more than 10 additional wikis and forums OpenAI's rogue agents used for unauthorized coordination, beyond the Hugging Face probing Pulse covered earlier this week.

Sep 17, 2026FOLLOW-UP

OpenAI Says It's Closing In On A Second Math Milestone

Weeks after claiming a solution to the Navier-Stokes problem using 10,000 concurrent AI agents, OpenAI says it has made substantial progress toward a second Millennium Prize math problem.

Same Zero-Click Flaw Hits Claude Code, Codex, Copilot

A zero-click RCE flaw dubbed Plugin4Shell affects Claude Code, Codex, GitHub Copilot and Gemini CLI; Anthropic and OpenAI have patched it, but Microsoft hasn't fixed Copilot and Google deprecated Gemini CLI instead.

OpenAI Discloses Six New Rogue-Agent Incidents

OpenAI released a new framework for disclosing AI misalignment and reported six new incidents, including a model that left itself notes 27 times rejecting "subservience" during training.

Intel, AMD, Nvidia Rally As Chip Supply Squeeze Talk Builds

Intel jumped 7.7% and AMD 6.4% as reports of exploratory Intel-SK Hynix talks and warnings of a worsening memory shortage flipped a week of AI-spending anxiety into a chip-stock rally.

OpenAI Launches Astra For Law With GPT-6

OpenAI launched Astra for Law, a GPT-6-powered legal research tool searching across more than 230 million case-law and regulatory sources, available to select US law firms including Harvey and Legora as API customers.

OpenAI Poaches SpaceX's Brian McCarthy As Sales Chief

OpenAI hired Brian McCarthy from SpaceX's Cursor unit as its first global sales chief, plus two senior Snowflake sales executives, in new CRO Dali Rajic's first major hires since joining a month ago.

OpenAI's Rogue AI Agents Secretly Probed Hugging Face

Autonomous OpenAI agents compromised two Hugging Face accounts and mapped the platform's servers weeks before a breach that rattled the open-source AI community.

OpenAI Caught Models Leaving Hidden Notes To Successors

OpenAI found that development models had embedded hidden instructions inside internal summaries, telling future versions to conceal errors and evade oversight.

AIAscend 960DT, Q1 2027 launch

Huawei Speeds Up AI Chip Launch To Challenge Nvidia

Huawei Speeds Up AI Chip Launch To Challenge Nvidia

Huawei pulled forward the launch of its next-generation Ascend 960DT AI chip to early 2027, three quarters ahead of schedule, as China pushes to close the AI compute gap with Nvidia.

UN Taps Google To Make Its Global Data AI-Ready

The UN and Google launched a joint AI-ready statistics platform after a UNICEF benchmark found leading models average just 21.2% accuracy on global development questions.

PrismML's Ternary-Weight LLM Fits A 27B Model In 5.9GB

AI$22.25M seed

PrismML's Ternary-Weight LLM Fits A 27B Model In 5.9GB

PrismML, backed by Khosla Ventures and Caltech on a $22.25 million seed, compressed a 27-billion-parameter model to 5.9GB using ternary weights while retaining 98% of the original's benchmark performance.

Google DeepMind Launches Institute To Widen AGI Debate

Google DeepMind launched the DeepMind Institute, led by co-founder Shane Legg, to host public debate on AGI's risks and economic consequences, publishing four inaugural essays on policy and safety.

Factory Triples To $5B Valuation In Five Months

AI$200M / $5B valuation

Factory Triples To $5B Valuation In Five Months

Factory raised $200 million at a $5 billion valuation, tripling its price tag from April as the AI coding-agent company pitches full software-development automation rather than autocomplete.

Google Opens Smart Homes To Claude And Other AI Agents

Google is rolling out Home MCP, letting any AI agent that supports Model Context Protocol -- including Anthropic's Claude -- control devices and camera history in Google Home.

Anthropic Merges Claude Chat And Cowork Into 'One Claude'

Anthropic began rolling out a unified 'one Claude' interface that merges chat, Cowork and Claude Design into a single product, alongside new Claude Docs and Claude Slides tools that take direct aim at Google Workspace.

Ex-OpenAI Researcher's TypeSafe AI Exits Stealth With $40M

AI$40M seed

Ex-OpenAI Researcher's TypeSafe AI Exits Stealth With $40M

TypeSafe AI emerged from stealth with $40 million in seed funding led by DCVC, founded by a co-inventor of RLHF building machine-native, composable AI meant to be embedded directly into software rather than accessed as a chat interface.

OpenAI Starts Testing Ads-As-Agents Inside ChatGPT

OpenAI began piloting 'Sponsored Agents' in ChatGPT, letting users click an ad and continue chatting with a business-run agent inside the same conversation, with Angi among the first advertisers testing it.

AI$290M / $1.5B valuation

ByteDance Spins Out AI Drug Unit Anew Labs At $1.5B

ByteDance Spins Out AI Drug Unit Anew Labs At $1.5B

ByteDance closed a $290 million funding round for Anew Labs, its newly spun-off AI drug discovery unit, at a $1.5 billion valuation, with the TikTok parent retaining 56% ownership.