Webpage: praxis-research.org/grafting | Paper link: https://arxiv.org/abs/2610.00767 Work done during Peter and Dani's MATS Fellowship (10.0) under the mentorship of Shi, with Jinghua Ou as Research Manager. TLDR; We share a simple modification to synthetic document fine-tuning (SDF) that works surprisingly well: do fine-tuning on a pre-training checkpoint and add the resulting weight difference onto the post-trained version that you want to deploy. We call this grafting. We empirically validate

AI can save small businesses time, reduce repetitive work, and unlock capabilities that previously required more people, but where should you actually start? In this episode, Matt and Mike break down how small businesses can introduce AI without getting overwhelmed, from identifying high-frequency, low-risk tasks and protecting sensitive business data to turning successful experiments into repeatable AI workflows. We also discuss why you shouldn't automate everything at once, how your existing

I'm here with a preview of Blockers, a new audiobook from Michael Lewis, the author of Liars Poker, Moneyball, The Big Short and more. In Blockers, Lewis tells the story of a handful of people who dedicated their careers to public service—until the Department of Government Efficiency decided they were a problem. These were people who believed they had a calling—people who kept American's sensitive tax data out of the wrong hands, made sure that politicians and civil servants played by

The author of The AGI Chronicles says his book “was written by a very tired, overworked, under-slept human being.”Join WIRED’s best and brightest on Uncanny Valley as they dissect the collision of tech, politics, finance, and business, from Alexis Ohanian’s newest tech venture to the effects of inaccurate information from artificial intelligence (AI) chatbots on social protests. Learn more about your ad choices. Visit podcastchoices.com/adchoices

The new scam that’s forcing people to call cyber criminals, Belkin’s new iPhone screen protectors that do more than just protect your screen, the telcos join forces to activate Temporary Disaster Roaming, Samsung has released new Galaxy Tab S12 tablets, we check out the Sennheiser HD 480 Pro Plus headphones, Snapmaker launches its new U1 3D printer in Australia and we’ll answer all your questions in the Tech Guide Help Desk.

Can you help me make more podcasts? Consider supporting me on Patreon as the service is 100% funded by you: https://EVne.ws/patreonYou can read all the latest news on the blog here: https://EVne.ws/blog Subscribe for free and listen to the podcast on audio platforms:➤ Apple: https://EVne.ws/apple➤ YouTube Music: https://EVne.ws/youtubemusic➤ Spotify: https://EVne.ws/spotify➤ TuneIn: https://EVne.ws/tunein➤ iHeart: https://EVne.ws/iheart KIA EXPANDS GT RANGE WITH EV3 GT, EV4 GT AND EV5 GT

Check out my new book AI Augmented Teams on Amazon or on my website paidar.ai/books.Regulated industries are under pressure to adopt AI without compromising security, compliance, or critical thinking. In this episode, Aaron Bach and Dr. Darren Pulsipher discuss why blanket AI bans fail, how shadow AI increases organizational risk, and what leaders can do to adopt AI responsibly.The conversation explores AI sovereignty, multi-model strategies, sensitive-data redaction, observability, human

Authors: Kyle O’Brien¹, Spencer Kitts², Cameron Tice¹, Alek Westover² ¹Geodesic Research, ²Redwood Research Figure 1: We pretrained LLMs from scratch with and without filtering subversion-relevant information, finding that subversion-relevant knowledge is significantly diminished while retaining general ML knowledge. TLDR: Filtering subversion-relevant information — subversion strategies, information about subversion defences, and empirical evaluations of the effectiveness of subversion strategies

Diet TBPN delivers the best of today’s TBPN episode in 30 minutes. TBPN is a live tech talk show hosted by John Coogan and Jordi Hays, streaming weekdays 11–2 PT on X and YouTube, with each episode posted to podcast platforms right after.Described by The New York Times as “Silicon Valley’s newest obsession,” the show has recently featured Mark Zuckerberg, Sam Altman, Mark Cuban, and Satya Nadella.TBPN is made possible by:Ramp - https://ramp.comPublic - https://public.comCisco -

This Week In Startups is made possible by:Vanta: https://www.vanta.com/twist DigitalOcean: https://do.co/twist Sentry: http://sentry.io/twist Plaud: http://Plaud.ai/twist Today’s show: The smarter AI gets, the more it just does the work for you, and the less helpful it becomes as a tutor. Shan Reddy of Aristotle joins Jason to demo a voice-first tutor built to run a session the way a $200-an-hour human tutor would. We also dive into why frontier models are trained to hand you answers rather than

AI can summarize documents, analyze spreadsheets, search internal systems, and automate entire workflows—but what happens to your sensitive data once it enters those systems?In this episode, David and Sophia explore the growing security risks created by rapid AI adoption and explain why traditional data protection tools may be unable to see where confidential information is actually going.From shadow AI and public chatbots to retrieval augmented generation, vector databases, AI agents, and

OpenAI helped ignite the generative AI boom—but what happens when the cost of building that future becomes larger than the business itself?In this episode, David and Sophia examine the financial and physical infrastructure behind the AI revolution, focusing on the enormous capital requirements, data center construction, energy demand, investor exposure, and systemic risks described in the source material.The conversation asks a provocative question: has the race to build increasingly

What happens when software developers stop writing code by hand and start managing AI agents instead?In this episode, David and Sophia explore a major shift happening across the software industry: the move from traditional programming toward AI-generated code, agentic workflows, and automated software factories.The discussion begins with David Heinemeier Hansson, creator of Ruby on Rails, and his claim that hand coding is no longer the normal course of business at 37signals. From there, the

(01:17) - AI Video Game Mods(18:04) - 𝕏 Timeline Reactions (37:13) - Chinese Cars Take Over Britain (43:49) - Fiat's $15k Microcar (57:09) - Sean Frank, CEO of Ridge, discusses the company’s rapid growth, his decision not to sell its business data to AI firms, and how AI is transforming e-commerce operations, advertising, and creative production. He also explores agent-driven shopping, TikTok marketing, consumer spending, celebrity partnerships, and Ridge’s potential expansion into

“You actually have the president and the Treasury working for Bitcoin because they’re working for a weaker dollar.” Matt Cole joins to discuss why he believes America’s debt crisis leaves the government with few good options, how rising bond yields could force the Treasury and Fed to intervene, and why a weaker dollar could fuel Bitcoin’s next bull market. We also get into why AI could create abundance without generating enough tax revenue to fix the debt problem, his case for $500K Bitcoin by

Welcome to the Bear Lair!* (No, it’s nothing like that other show.) As we continue to explore the world of AI tools and vibe coding, we invited a few enterprising Vergecast fans to share their projects — and we invited The Verge’s Hayden Field to judge them with us. We’ll see what people are working on, and whether any of them will make us fabulously wealthy.*We came up with this name before we became aware that someone else thought of it first... no relation to the APTNtv show Bears'

This article was created by Forethought. See all of our research on our website. How persuasive are present-day AI systems? Unfortunately, it is surprisingly difficult to know where AI capabilities in persuasion are today (September 2026), or even three months ago. Benchmarking AI capabilities well is always hard, but I think humanity's understanding of AI's persuasive abilities is substantially worse than our understanding of, say, AI cybersecurity or math abilities. Still, this question is worth

Reflection is soon releasing Beam, its first open-weight AI model, in a bid to become the DeepSeek of the West. Trained from scratch, Beam is designed for coding, reasoning, and AI agents. Reflection's benchmarks put it alongside the strongest open models, with substantially more efficient token economics.To mark Beam’s debut, I'm joined by Reflection's co-founders: CEO Misha Laskin, who led reward modeling for Google’s Gemini, and CTO Ioannis Antonoglou, who helped create AlphaGo and AlphaZero

An ex-OpenAI safety staffer quit, saying AI labs needed to get it right the first time. Plus, Western open-weight models were coming, Meta's Muse kept a page on everyone you know, SpaceXAI got renamed SpaceXSI, and Apple tightened Full Disk Access.David Robinson, ex-OpenAI safety and policy: SV lacks a safety-centric culture; labs must study other fields' safety approaches; time for trial and error's over (The Atlantic) OpenAI safety employee David Robinson quits, says AI firms should be run

Oil flows through the Middle East are back near pre-war levels, but 10 year yields and the MOVE index are telling a different story.🔗 https://bitcoinproducts.com In this episode: Why Middle East crude flows are back to pre-war levels while diesel and refined products still lag What the US and China statement on Iran and Hormuz tolls signals How the push for a China-free rare earth supply chain is going Why the MOVE index and French bond spreads are blowing out How stablecoins and Treasury bills