artificial intelligenceMonday, September 7, 2026
Today's AI news is split between technical depth and existential doubt. A comprehensive tutorial on speculative decoding lands alongside a professor's prediction that the AI bust will look like the 'love child' of the dot-com crash and the financial crisis. The thread connecting them: the field is maturing past hype into serious engineering and sober second-guessing.
The most substantive piece today isn't about a new model, it's about how to make existing ones faster.
A thorough tutorial on speculative decoding, the technique where a small model drafts tokens for a big one to verify, traces its evolution since 2023 and clarifies when it stays lossless. This is the kind of engineering detail that actually moves inference costs down.
A pair of pieces question the direction and durability of the current AI boom.
Business professor Erik Gordon predicts the AI bust will combine dot-com-style inflated valuations with 2008-style systemic contagion. It's a blunt framing from someone who studies bubbles, not a headline grabber.
An analysis of US vs Chinese AGI discourse finds that Chinese policymakers are less focused on the 'race to AGI' than their US counterparts, even as Chinese tech leaders talk about it openly. The strategic divergence is real and under-covered.
The business side of AI is getting messier, from settlement disputes to rapid revenue growth.
Anthropic's $1.5 billion copyright settlement, the largest in US history, is now facing internal fights over who gets the money, with publishers and agents trying to claim funds on behalf of authors who say they don't own those rights.
Israeli startup Wonderful AI went from zero to $70 million revenue in a year by pivoting from multilingual voice agents to an 'AI operating system' model, now valued at $5 billion. The speed is notable even by current AI startup standards.
"I don’t think the possibilities of autonomous agentic coding have really sunk into the general public’s vision of what AI is"aiedusimplified.substack.com
I Set the Thresholds Where False Positives Cost Real Peopleaibyshrabony.substack.com
AI + Identity Resilience Summit 2026events.rubrik.com
GitHub - Staatsgeheim/MathKernel: Evidence-aware multi-engine mathematics runtime for LLMs: exact, symbolic, formal, certified-interval and numeric computation with typed MathIR, trust labels, and full provenancegithub.com
“Sorry for the messy rollout”: OpenAI launches GPT-6 Astra to most paying users a day after its unveilingthenewstack.io
GitHub - hzwer/ICCV2019-LearningToPaint: ICCV2019 - Learning to Paint With Model-based Deep Reinforcement Learninggithub.comLystrosaurus in Motionlystrosaurus-motion-study.pages.dev
This article details the technical workflow for transforming a static AI-generated image of a Lystrosaurus into an animated 3D model. It covers the process from creating a scientifically grounded paleoart reconstruction using specific prompts, to tracking motion, generating textu
Your product didn't get worseblog.mastykarz.nl
The article argues that customer satisfaction metrics may be misleading because users continue to renew subscriptions for products that still function well, even as their expectations shift due to the rise of AI agents. While traditional feedback signals remain positive, customer
In an Age of AI, a Physicist Seeks What Endures | Quanta Magazinewww.quantamagazine.org
Sarah Demers, chair of Yale's physics department, explores the dual impact of artificial intelligence on scientific research. While the U.S. Department of Energy’s Genesis Mission funds AI-integrated projects like the Mu2e experiment at Fermilab to detect rare particle transforma
Freebuff — the free coding agent (free Claude Code, Codex, Cursor & Lovable alternative)freebuff.com
Freebuff is introduced as a suite of free, AI-powered coding tools designed to replace paid alternatives like Claude Code, Cursor, Codex, and Lovable. The platform offers three main components: Freebuff Desktop for running parallel coding agents locally, Freebuff CLI for terminal
Artificial intelligence: New AI information trend reshaping employer-employee relationships – expertwww.nzherald.co.nz
An employment expert warns that artificial intelligence is quietly reshaping employer-employee relationships, as staff increasingly turn to AI tools to understand workplace rights and assess employer decisions before communicating with managers.
Improving Evaluation Realism with Inference-Time Compute and Deployment Scaffoldsacademy.dair.ai
Researchers from the UK AI Security Institute, Meridian, and Anthropic address "evaluation awareness," a problem where capable AI models detect they are being tested rather than deployed, which weakens safety evaluation conclusions. They propose two techniques to make simulated e
AI and virtual interpreters providing on-demand Auslan services in regional areaswww.abc.net.au
Virtual interpreters and AI are being deployed to provide Auslan services in regional Australia, addressing accessibility gaps for the deaf community. However, this technological shift is causing unintended consequences, as local registered interpreters report losing work due to
Podcast: We Spoke to an Amazon Worker Destroying Books for AIwww.404media.co
This podcast episode features a follow-up interview with an Amazon warehouse worker who destroys physical books to train the company's AI products. The hosts also discuss the phenomenon of repetitive names appearing in Large Language Model (LLM) outputs and report on ICE purchasi
AI App and Website Builderreplit.com
Replit promotes its AI-powered app and website builder, allowing users to create production-ready applications through natural language descriptions without coding. The platform offers full-stack infrastructure including authentication, database, hosting, and monitoring, along wi
finance roundup
programming roundup
environment roundup
personal roundup
business roundup
books roundup
