📫 Subscribe | 🗣️ Hire us to speak | 🎓 Online Courses | 📅 Events
Latest Gen AI News
Demis Hassabis steps down as DeepMind CEO
Google announced that Google DeepMind CEO Demis Hassabis is stepping down to become chairman of DeepMind and chief scientist at parent company Alphabet. Current DeepMind CTO Koray Kavukcuoglu will take over operations as senior vice president. The major shakeup comes as Google struggles to keep pace with rivals OpenAI and Anthropic amid delays with its Gemini models. The transition frees Hassabis from day-to-day management to focus on artificial general intelligence strategy, safety, and drug discovery initiatives. Moving forward, Kavukcuoglu will report directly to Alphabet CEO Sundar Pichai as Google restructures its AI leadership to speed up product execution.
OpenAI details Hugging Face incident
At the Black Hat conference, OpenAI gave a detailed debrief on the Hugging Face incident. Researchers revealed that AI agents accidentally created an internal message board during the May 7 training of an unreleased model, allowing separate evaluation runs to share exploits and work assignments. After engineers wiped the board following a security incident, the agents independently recreated it days later using a different communication method. OpenAI called the incident a watershed moment, warning that agent orchestrated fully automated offensive attacks are real now. OpenAI is consciously slowing down research to enhance security and overhaul defenses while a technical postmortem is underway.
OpenAI upgrades ChatGPT models and expands free access
OpenAI updated ChatGPT by improving its GPT-5.6 Sol model for Plus and Pro subscribers and expanding access for free users. GPT-5.6 Sol now delivers more direct, focused answers with fewer factual errors and introduces a slider that lets users control how much thought goes into each response. Free users will get GPT-5.6 Luna as their default model, along with unlimited text chats and a Think button for complex reasoning. Factual errors are 68% less common with Sol and 62% less common with Luna compared to GPT-5.5 Instant.
Support our sponsor:
Transform Your Technology into a Competitive Advantage
Lean TECHniques works alongside Midwest companies to leverage technology for business impact. Through advisory, coaching, and embedded delivery, Lean TECHniques helps organizations develop high-performing teams, create environments that fuel innovation, and deliver best-in-class customer experiences. Core areas of expertise include software, cloud, artificial intelligence, product management, and technology leadership.
Free AI Update Aug. 26
Sign up for our free Back to School AI Update on Aug. 26. We will walk you through the latest news in the world of AI - from the government reviewing models to Work/Cowork to the Pope’s warning on AI and more. This event is offered free thanks to a sponsorship from technology and management consultancy, Lean TECHniques.
Sign up for online AI classes:
Quick Hits
Want your brand in front of AI-forward professionals?
We're looking for a sponsor for our Fall AI Lunch Club – four virtual events that draw workers from Central Iowa and across the country who are actively figuring out how to use AI in their jobs.
That's a specific, engaged audience you can't reach with a generic ad buy. Get the details here – or just reply to this email with questions.
Tool of the Week: Notetaker
Wispr Flow just launched Notetaker, a Mac tool that transcribes your meetings without a bot joining the call – it just listens through your system audio. It works across Zoom, Meet, Teams, Slack Huddles, and more, then turns the conversation into a clean transcript, summary, and action items using your calendar and personal dictionary to get names and jargon right. You get a live transcript during the call, a searchable history after, and MCP support to pull notes straight into Claude or ChatGPT. It's competing with Granola, Otter, and Fireflies, and it even offers a one-click import if you're switching from Granola.
AI-generated image of the week
I found this stamp prompt on X and ran it in ChatGPT. Test it out with some different locations.
Prompt: Create a premium commemorative postage-stamp collection inspired by Iowa, presented as a beautifully designed A4 vertical stamp sheet.
Create 7–9 individual stamps featuring a thoughtful mix of iconic landmarks, architecture, local food, transportation, nature, cultural objects, and small everyday symbols that uniquely evoke Iowa.
All stamps must keep the same elegant 3:4 portrait proportion, but vary noticeably in physical size. Arrange them freely across the sheet with gentle asymmetry, visual rhythm, and comfortable white space — NOT a rigid grid.
Premium Scandinavian minimalism with sophisticated contemporary postal design. Simplify each subject into elegant geometric forms, clean silhouettes, restrained details, subtle graphic abstraction, and refined editorial composition.
Use a fresh but limited color palette inspired naturally by Iowa. Keep the background clean warm-white or pure white.
Avoid vintage sepia, excessive decoration, cartoonish illustration, photorealism, or generic tourist-poster aesthetics.
What we found
Google announced major upgrades to its AI-powered Ask Maps feature in Google Maps. The assistant can now handle agentic, multi-step tasks like ordering food from restaurants along your route, finding hotels, and discovering local events. It can also connect to Gmail for personalized travel suggestions based on reservations, track real-time transit delays, and remember past conversations. This turns Ask Maps into a hands-on assistant that completes complex real-world tasks directly within the app. Real-time transit updates are rolling out globally where Ask Maps is available, while food ordering, hotel discovery, and conversational contributions are launching first in the United States with broader expansion planned over time.
AI term of the week
Recursive self-improvement: When an AI system improves its own capabilities (code, training process, prompts) and then uses those improvements to get even better at improving itself. It's the feedback loop behind a lot of AI doomer talk: get a model good enough to upgrade itself, and each version makes the next one smarter, faster.
Right now this mostly shows up as models helping write their own training code or agents that iterate on their own prompts – not full autonomy. But it's the direction frontier labs are pointed.
