AI is great. I guess.
Sep 23, 2026
•
4 min read
The price race is cutting frontier AI costs nearly in half, while Qualcomm is putting agents on-device, Epoch AI is tracking the plunging cost of thought, and DeepSeek is betting on Huawei chips.
Sep 22, 2026
5 min read
PLUS: Frontier-model rumors, Meta’s Muse outperforming ChatGPT, Alibaba’s AI chips plan, and OpenAI’s frontier standards.
Sep 21, 2026
PLUS: China is pushing DRAM into mass production, Samsung is scaling HBM4, and a graph world model shows how to make long AI workflows more reliable.
Sep 19, 2026
PLUS: Hype around Jev points to AI as a new decision layer, while Anthropic and Accenture commit $2 billion to evaluation and OpenAI faces a projected $278 billion cash burn by 2030.
Sep 18, 2026
PLUS: Huawei scales AI clusters. Z.ai uses GLM-5.3 to build its inference stack. OpenAI’s safety framework and Qwen’s multimodal agents.
Sep 17, 2026
PLUS: The race to build voice agents, OpenAI’s $40M health-data push, and a smarter way to manage long-running agents.
Sep 16, 2026
PLUS: GPT-6 speculation leads the week, Anthropic’s 2.16GW Australia bet, Gemini 3.8 Live, and AI agents learning to govern themselves.
Sep 14, 2026
OpenAI’s safety push, 4-hi HBM economics, China’s BRICS AI plan and the market’s latest warnings reveal what comes next for the industry.
Sep 13, 2026
PLUS: China is shifting from model-building to AI agents, while a rogue swarm hits RubyGems and OpenAI scales Habitat past 20M requests per second.
Sep 11, 2026
PLUS: GPT-Live-1 brings voice agents to the API, Anthropic tracks Claude misuse at scale, and a new paper rethinks how multiple models work together.
Sep 10, 2026
PLUS: HBM shortages are pushing Chinese chip prices higher, Anthropic maps three economic futures, and agent safety is becoming infrastructure.
Sep 9, 2026
PLUS: OpenAI claims a Navier–Stokes proof, expands chip research with Samsung, and DeepSeek’s fundraising sparks a costly shadow market.
Sep 8, 2026
3 min read
PLUS: TPU economics put pressure on CUDA, ChatGPT Work learns from connected work data, and AI deployment moves from raw speed to quality, supervision and reusable systems.
Sep 7, 2026
PLUS: OpenAI’s 3.1 agent-workday ratio, Jensen Huang’s AGI claim, and NVIDIA’s reported $2.5B Thinking Machines investment.
Sep 6, 2026
PLUS: OpenAI is rethinking agent disclosure as South Korea builds the physical base for sovereign AI.
Sep 5, 2026
New model capability is colliding with weaker observability, competing AGI definitions and a growing agent-safety problem.
Sep 4, 2026
PLUS: A $13B power shift is reshaping open models, OpenAI is putting $1B behind cyberdefense, and local AI is moving onto more machines.
Sep 3, 2026
PLUS: Meta’s Muse Spark 1.3 forgets less, Fable tops WebDev Arena, Grok Bot becomes a 24/7 AI team, and a GPT-6 branch sparks speculation.
Sep 2, 2026
6 min read
TODAY: OpenAI built a model that finds zero-days; Anthropic cuts agent costs, Google closes the coding gap; and AI infrastructure races toward $32T.
Sep 1, 2026
PLUS: Hacker-Opus exposes reward hacking. ChatGPT’s ads hit a $1B annualized run rate. Coding agents go multi-session, and AI infrastructure moves into the real economy.
Aug 31, 2026
OpenAI’s Macs for AI agents, SpaceX’s turbine-pollution tradeoff, Jalapeño’s latency bet, and frontier AI cyber risk.
Aug 30, 2026
Beyond Nvidia: AI compute's problem, OpenAI is rethinking inference with Jalapeno, Astra is taking shape, and Agent Societies are learning to coordinate through persistent worlds.
Aug 29, 2026
PLUS: GLM-5.3 opens up, then adds a $10B security gate. Anthropic says Claude reduced 10 alignment-failure categories.
Aug 28, 2026
PLUS: Anthropic’s hardware push, Hunyuan 4, Thinking Machines’ new Text-to-SQL method, and a 100+ group cyberattack warning.
Aug 27, 2026
PLUS: Anthropic opened 250,000 Claude conversations to outside researchers, while AWS and NVIDIA announced 2 million more GPUs for the AI race.
