- AI Fire
- Posts
- 💭 Google DeepMind Rebuilds a Love Story
In partnership with
Google DeepMind helped bring a couple’s first meeting to the screen more than 70 years later, even though no camera captured it. One small correction from Ethelle shows where an AI-made memory gets complicated.
What's on FIRE 🔥
IN PARTNERSHIP WITH BELAY
As your business grows, your role should evolve with it. But for many leaders, it doesn’t. They stay buried in tasks, decisions, and responsibilities they’ve already outgrown.
BELAY created the free resource From Operator to Owner to show leaders how to step out of the day-to-day without losing control and what it takes to operate at the next level. At BELAY, we match you with U.S.-based Assistants who take on the operational work that’s keeping you stuck.
AI INSIGHTS
Burt and Ethelle Shatz have been married for over 70 years. As Burt’s memories fade, one moment has become harder for him to recall: the day they met at a student co-op in Cleveland. No one took a photo or filmed it.
For the short documentary Love, Rendered, filmmakers worked with Google DeepMind to recreate parts of the couple’s past. Ethelle helped guide the work, even correcting details like the curve of a staircase and the shape of a shoe heel.
The team used AI in two ways:
Old photos became references. Generative models restored pictures of Burt and Ethelle when they were young, helping the team keep their younger faces recognizable.
Present-day gestures shaped the performance. The team mapped small details, including Burt’s head tilt and expressions, onto the recreated scenes.
The result is a moving scene of a day no camera captured. It’s also a reconstruction, guided by Ethelle’s memories rather than a record of exactly what happened.
If AI can show us moments we never filmed, how do we keep the line clear between a memory and a scene we’ve created from it?
PRESENTED BY HUBSPOT
The Future of AI in Marketing. Your Shortcut to Smarter, Faster Marketing.
Unlock a focused set of AI strategies built to streamline your work and maximize impact. This guide delivers the practical tactics and tools marketers need to start seeing results right away:
7 high-impact AI strategies to accelerate your marketing performance
Practical use cases for content creation, lead gen, and personalization
Expert insights into how top marketers are using AI today
A framework to evaluate and implement AI tools efficiently
Stay ahead of the curve with these top strategies AI helped develop for marketers, built for real-world results.
AI SOURCES FROM AI FIRE
1. How to Turn Claude Into Your One-Person Marketing Team in Under 1 Hour (Part 2: Let It Run). In Part 1, I built the "brain," a Claude Code project with brand context, customer information, and approved assets. In Part 2, I actually put it to work. Campaign planning comes first, then prompts, then production.
2. This Claude Code Agentic OS is the New Meta: Voice Control, Skills, and a Second Brain. Something interesting happens when Claude Code gets access to a structured Obsidian vault. This setup may be one of the most interesting Claude workflows to watch right now.
3. OpenClaw 2.0 is INSANE: Better, Free & UNLIMITED AI Assistant in Mins (Full Setup Guide). This is a massive, ground-up update. You’ll see the complete setup, what makes OpenClaw 2.0 different, and how quickly you can get your own assistant running.
FIRE RECAP: BIGGEST AI NEWS THIS WEEK
🧮 OpenAI says about 10,000 AI agents found a proof for the famous Navier–Stokes math problem, then GPT‑6 Astra checked it in Lean. The Clay Mathematics Institute says the problem has apparently been settled, but its review process is still ahead.
😟 Former Anthropic and OpenAI researcher Jacob Coxon resigned and warned that AI labs are moving too fast toward self-improving systems. That’s his assessment. He told Axios he left before his Anthropic equity vested.
🛑 Anthropic CEO Dario Amodei is calling for a slower AI race. His plan would give independent safety reviewers ongoing access inside AI labs and push companies and governments to coordinate.
🤖 GPT‑6 Astra cleared all 48 levels of an “I’m Not a Robot” game. Fun demo, but real sites are tougher: in one seven-site test, Astra completed two signups, and neither had a CAPTCHA.
🧬 Google DeepMind released AlphaGenome Atlas, a searchable map of predictions for 9 billion possible DNA-letter changes. Researchers can use its impact score to decide which changes to study first.
TODAY IN AI
AI HIGHLIGHTS
📉 OpenAI’s Sam Altman says a 2026 IPO would be ill-timed as AI safety concerns grow. In a new Fortune interview, he said the company won’t go public until 2027.
📱 California has banned social platforms from showing autoplay and personalized feeds to users under 16. The new laws also require AI companion chatbots to undergo child safety audits and yearly risk checks.
🎵 Universal Music Group and ElevenLabs are building an AI music platform where fans can remix tracks from participating artists. Their multi-year deal covers licensed music, but the platform is still in development.
🛒 Amazon Ads is bringing its advertisers into ChatGPT through a new pilot with OpenAI. Selected U.S. brands, including Delta Vacations, are testing ads that appear as people explore options and make decisions in chat.
🛑 The UK government has rejected a proposed emergency AI “kill switch.” It says blocking a model in the UK wouldn’t stop that model from being developed or misused elsewhere. The proposal can still move through Parliament.
💰 AI Daily Fundraising: Sequoia Capital is close to leading a new round for Mecka AI at a $500M valuation, just three months after the startup raised $60M. Mecka pays people to record everyday tasks with phones and body sensors, creating real-world data to train humanoid robots.
Even when those documents are 200-page engineering specs, scanned tables, handwritten notes, and drawings. During our webinar we'll cover:
How complex documents become structured, traceable data without losing what makes them complex
Why agents fail when the data layer is wrong, and how to know before it ships
A real workflow from raw documents to production agents, with York IE
NEW EMPOWERED AI TOOLS
🎯 Youkti remembers every account, conversation, and deal, then tells sales teams which prospects are ready to buy and what action to take next.
🔌 Monid connects AI agents to 1,800+ APIs for SEO, lead gen, social media, market data, video, and more through one key.
🎬 Clipto MCP lets Claude, ChatGPT, and other AI agents search local video, photos, and audio to find clips, build rough cuts, and source footage.
📅 Hey Noah is a proactive AI executive assistant that manages calendars, relationships, and follow-ups across email, SMS, and WhatsApp.
📣 AdAnt AI uses creative agents to plan, create, and iterate social ads for TikTok, Instagram, and YouTube.
AI BREAKTHROUGH
Google just introduced ToolGrad, a new framework for generating better training data for AI agents that use tools and APIs.
Traditional methods create a user prompt first, then make an AI search for the correct tool-use path, which often fails. ToolGrad reverses that process by building a verified tool-use chain first, then generating the matching user request. Main findings:
This “answer-first” approach achieved a 99.8% pass rate while producing longer tool-use workflows at lower cost.
Google used just 500 generated examples to fine-tune Gemma 3 models.
ToolGrad-12B scored 83.1 on the Berkeley Function Calling Leaderboard, roughly matching Gemini 2.5 Pro at 83.2 and beating Claude 4.5 Opus at 82.8 and GPT-5 at 74.4 in Google’s reported comparison.
Interestingly, the training data was generated by Gemini 2.5 Flash-Lite, yet the resulting Gemma model could outperform its own teacher on tool use.
ToolGrad shows that improving AI agents may depend heavily on how their training data is created. It starts with a working solution, now Google can generate reliable tool-use data much more efficiently.
We read your emails, comments, and poll replies daily
How would you rate today’s newsletter?Your feedback helps us create the best newsletter possible |
Hit reply and say Hello – we'd love to hear from you!
Like what you're reading? Forward it to friends, and they can sign up here.
Cheers,
The AI Fire Team
Reply