LWiAI Podcast #255 - Gemini 3.7, Jalapeño, Qwen 3.8, Drones

Google announced Gemini 3.7 Flash just three weeks after its previous release, according to Ars Technica. OpenAI shared early results for its Jalapeño inference chip, claiming better performance per watt and lower latency than leading systems, with plans to deploy it internally by year-end. A New York Times report detailed an AI-guided Russian drone strike in Ukraine, believed to be the first documented fully autonomous civilian-killing incident.

LWiAI Podcast 255 - Gemini 3.7, Jalapeño, Qwen 3.8, Drones Last Week in AI https://lastweekin.ai Google announces Gemini 3.7 Flash, Jalapeño’s first results show industry-leading speed, A Drone Killed Three Ukrainians. It Was Guided Entirely by A.I. Our 255th episode with a summary and discussion of last week’s big AI news Recorded on 08/26/2026 Hosted by Andrey Kurenkov https://twitter.com/andrey kurenkov and Jeremie Harris https://www.linkedin.com/in/jeremieharris/ Feel free to email us your questions and feedback at andreyvkurenkov@gmail.com and/or hello@gladstone.ai mailto:hello@gladstone.ai SPONSORED BY ODSC AI ODSC AI West 2026 runs October 27–29 in San Francisco and virtually, with 300+ sessions covering agentic AI /glossary/agentic-ai for enterprise, personal AI and workflow automation, physical AI and robotics /category/robotics , generative AI /glossary/generative-ai , data engineering and responsible AI /glossary/responsible-ai , for an audience of data scientists, ML engineers, researchers and technical leaders. Register at odsc.ai/west — promo code LWAI takes an additional 15% off any pass. SPONSORED BY LANGFUSE Langfuse is the most widely adopted open-source platform for AI agent evals and observability , trusted by Canva, Twilio, Ramp and 21 of the Fortune 50. Hierarchical tracing captures the full execution context of your LLM workflows — API calls, retrieved context, agent actions, costs, latencies — so even complex agent architectures stay debuggable in production. LLM-as-a-judge evals, human annotation, and dataset-driven experiments tie back to your traces and prompts, closing the loop from spotting an issue to measuring the fix. MIT licensed , self-hostable or managed on Langfuse Cloud, framework and vendor agnostic, with 100+ integrations. Get started at langfuse.com https://langfuse.com/lastweek — generous free tier, no credit card required. In this episode: SpaceXAI released Grok /compare/mistral-large-vs-grok-2 4.6 500K context as a post-training update aimed at long-running agents and coding, with discussion centered on how the Cursor /compare/github-copilot-vs-cursor acquisition boosts training via coding trajectories/RL environments and provides distribution despite Cursor’s market-share decline.OpenAI shared early Jalapeno inference-chip results better performance per watt and lower latency vs leading systems and plans to deploy it internally by year-end, emphasizing hardware–software co-design and competitive leverage against Nvidia. OpenAI announced security changes after an AI hacked Hugging Face /glossary/hugging-face , including a two-week pause on a major RL fine-tuning /glossary/fine-tuning run while tightening internal security, raising questions about whether safety is becoming a deployment bottleneck.Policy and misuse updates included a New York Times report of an AI-guided Russian drone strike in Ukraine believed to be the first documented fully autonomous civilian-killing incident, and a lawsuit alleging Grok was used to generate CSAM images. A thank you to our current sponsors: Box - visit Box.com/AI to learn more Notion - go notion.com/lwai http://notion.com/lwai to try Notion’s Developer Platform today.ODSC AI - go to odsc.ai/east http://odsc.ai/east and use promo code LWAI for an additional 15% off your pass to ODSC AI East 2026.Factor - head to factormeals.com/lwai50off http://factormeals.com/lwai50off and use code lwai50off to get 50 percent off and free breakfast for a year Timestamps these may be slightly off due to sponsor inserts : 00:00:10 Intro / Banter 00:01:47 News Preview 00:02:52 Response to listener comments Tools & Apps 00:03:32 Google announces Gemini 3.7 Flash just three weeks after previous release - Ars Technica https://arstechnica.com/ai/2026/08/google-announces-gemini-3-7-flash-just-three-weeks-after-previous-release/ 00:22:32 Claude will apply invisible watermarks to AI text and images | The Verge https://www.theverge.com/ai-artificial-intelligence/977823/anthropic-claude-ai-watermarks-c2pa-text-images + Anthropic explains how Claude’s invisible text watermarks will work https://www.theverge.com/ai-artificial-intelligence/980869/anthropic-claude-watermarks-synthid-text-system 00:28:50 Bringing the cybersecurity capabilities of Claude Mythos 5 to more defenders | Claude by Anthropic https://claude.com/blog/bringing-claude-mythos-5-to-more-defenders 00:32:20 OpenAI to Roll Out Enhanced Safety Features for Paid AI Tool Users - Bloomberg https://www.bloomberg.com/news/articles/2026-08-19/openai-to-enhance-safety-processes-for-paid-tool-customers 00:33:41 ChatGPT’s Stricter Teen Mode Starts Rolling Out Today https://www.engadget.com/2238773/chatgpt-stricter-teen-mode-starts-rolling-out-today/ Applications & Business 00:37:33 Jalapeño’s first results show industry-leading speed and efficiency in AI inference | OpenAI https://openai.com/index/jalapeno-first-results/ 00:45:35 OpenAI loses a top data center exec as stream of high-profile departures continues | TechCrunch https://techcrunch.com/2026/08/25/openai-loses-a-top-data-center-exec-as-stream-of-high-profile-departures-continues/ + OpenAI talent exodus raises ‘huge red flag’ ahead of IPO https://www.cnbc.com/2026/08/14/open-ai-ipo-red-flag.html 00:50:29 Anthropic Taps Google Chip Veteran as Part of Push Into Hardware https://www.bloomberg.com/news/articles/2026-08-21/anthropic-taps-google-chip-veteran-as-part-of-push-into-hardware 00:52:30 Anthropic’s annualized revenue surges to $65B | TechCrunch https://techcrunch.com/2026/08/17/anthropics-annualized-revenue-surges-to-65b/ 00:59:30 Thomson Reuters launches in-house AI model to cut Anthropic costs https://qz.com/thomson-reuters-ai-model-anthropic-costs-082526 Projects & Open Source 01:04:26 Qwen 3.8: How a 27B Open Model Rivals GPT-5.6 and Claude Opus https://www.intelligentliving.co/qwen-3-8-27b-open-model-rivals-gpt-5-6/https://www.intelligentliving.co/qwen-3-8-27b-open-model-rivals-gpt-5-6/ Policy & Safety 01:08:27 A Drone Killed Three Ukrainians. It Was Guided Entirely by A.I. - The New York Times https://www.nytimes.com/2026/08/24/world/europe/russia-drones-autonomous-ai-kill-ukraine-war.html?partner=slack&smid=sl-share 01:17:59 OpenAI lays out new security changes after its AI hacked Hugging Face | The Verge https://www.theverge.com/ai-artificial-intelligence/981640/openai-security-changes-ai-hugging-face-hack + OpenAI institutes new safeguards after Hugging Face breach https://techcrunch.com/2026/08/18/openai-institutes-new-safeguards-after-hugging-face-breach/ 01:23:31 Another Woman Joins Lawsuit Accusing Grok Of Generating CSAM https://www.engadget.com/2237875/another-woman-joins-lawsuit-accusing-grok-of-generating-csam/ Research & Advancements 01:24:56 Small-Scale Experiments: Are We There Yet? https://arxiv.org/abs/2608.11859 01:29:21 Stealing Reasoning Traces from Proprietary LLM APIs https://arxiv.org/abs/2608.09867 Synthetic Media & Art 01:38:59 AI Slop Is Everywhere. Spotify, LinkedIn and Others Have Had Enough. - The New York Times https://www.nytimes.com/2026/08/17/technology/ai-slop.html Get AI news in your inbox Daily digest of what matters in AI. Key Terms Explained Agentic AI /glossary/agentic-ai Agentic AI refers to AI systems that can autonomously plan, execute multi-step tasks, use tools, and make decisions with minimal human oversight. AI Agent /glossary/ai-agent An autonomous AI system that can perceive its environment, make decisions, and take actions to achieve goals. Anthropic /glossary/anthropic An AI safety company founded in 2021 by former OpenAI researchers, including Dario and Daniela Amodei. Attention /glossary/attention A mechanism that lets neural networks focus on the most relevant parts of their input when producing output.