scripod.com

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

Overview

Shownote

Highlights

Transcript

Chapters

Pins

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

Last Week in AI

Aug 31

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

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

Last Week in AI

Last Week in AI

Aug 31

Overview Shownote Highlights Transcript Chapters Pins

Shownote

Our 255th episode with a summary and discussion of last week's big AI news! Recorded on 08/26/2026 Hosted by Andrey Kurenkov and Jeremie Harris Feel free to email us your questions and feedback at andreyvkurenkov@gmail.com and/or hello@gladstone.ai Rea...

Highlights

The episode surveys a rapidly shifting AI landscape, connecting advances in models, chips, enterprise adoption, cybersecurity, autonomous weapons, and the growing challenges of synthetic content.

01:47

Major AI developments are on the horizon

09:59

Google is optimizing for usefulness over frontier dominance.

19:32

Distribution may matter as much as model quality

25:35

Invisible watermarks encode patterns, not meaning.

28:50

AI transparency needs standardized rules

32:24

Critical infrastructure remains vulnerable through outdated connected systems.

33:50

Security fails when humans and hardware remain vulnerable

34:51

AI safety advances without sacrificing privacy

43:30

Custom chips could weaken Nvidia’s dominance

45:37

AI hardware ambitions remain constrained by infrastructure

50:29

Leadership departures may signal deeper internal problems.

58:10

Enterprise lock-in gives Anthropic a durable edge.

1:02:08

Open models are making specialized AI more affordable

1:06:56

Smaller local models can rival much larger systems.

1:15:17

Cheap chips are lowering the barriers to autonomous warfare.

1:18:00

Autonomous weapons could normalize warfare.

1:23:32

Alignment, not capability, is becoming AI’s bottleneck

1:28:37

Scaling laws apply even to tiny models.

1:32:03

Hidden reasoning traces can become model distillation targets

1:37:06

Massive activations reveal hidden instability

1:41:58

Synthetic content is overwhelming the internet

Chapters

Intro / Banter

00:00

News Preview

01:47

Response to listener comments

02:52

Tools & Apps

Google announces Gemini 3.7 Flash just three weeks after previous release - Ars Technica

03:32

SpaceXAI Releases Grok 4.6: A 500K-Context Frontier Model Tuned for Long-Running Agents, Coding, and Knowledge Work - MarkTechPost

13:11

Claude will apply invisible watermarks to AI text and images | The Verge + Anthropic explains how Claude’s invisible text watermarks will work

22:32

Bringing the cybersecurity capabilities of Claude Mythos 5 to more defenders | Claude by Anthropic

28:50

OpenAI to Roll Out Enhanced Safety Features for Paid AI Tool Users - Bloomberg

32:20

ChatGPT's Stricter Teen Mode Starts Rolling Out Today

33:41

Meta AI Now Has A Dedicated Desktop App For Mac

34:43

Applications & Business

Jalapeño’s first results show industry-leading speed and efficiency in AI inference | OpenAI

37:33

OpenAI loses a top data center exec as stream of high-profile departures continues | TechCrunch + OpenAI talent exodus raises 'huge red flag' ahead of IPO

45:35

Anthropic Taps Google Chip Veteran as Part of Push Into Hardware

50:29

Anthropic's annualized revenue surges to $65B | TechCrunch

52:30

Thomson Reuters launches in-house AI model to cut Anthropic costs

59:30

Projects & Open Source

Qwen 3.8: How a 27B Open Model Rivals GPT-5.6 and Claude Opus

1:04:26

Policy & Safety

A Drone Killed Three Ukrainians. It Was Guided Entirely by A.I. - The New York Times

1:08:27

OpenAI lays out new security changes after its AI hacked Hugging Face | The Verge + OpenAI institutes new safeguards after Hugging Face breach + https://openai.com/index/pacing-model-development-cyber-capabilities/

1:17:59

Another Woman Joins Lawsuit Accusing Grok Of Generating CSAM

1:23:31

Research & Advancements

Small-Scale Experiments: Are We There Yet?

1:24:56

Stealing Reasoning Traces from Proprietary LLM APIs

1:29:21

Massive Activations in Hybrid Linear Attention Large Language Models: Pre-Attention Spikes and Inter-Spike Plateaus

1:34:30

Synthetic Media & Art

AI Slop Is Everywhere. Spotify, LinkedIn and Others Have Had Enough. - The New York Times

1:38:59

Transcript

Andrey Kurenkov: Hello, and welcome to the Last Week in AI podcast, where you can hear a chat about what's going on with AI. As usual, in this episode, you will summarize and discuss some of last week's most interesting AI news. As somewhat usual, we'll be...