LAST DAYS TO EARN UP TO 25% BONUS SHARES — INVEST IN US BY AUG 28

Welcome, AI Visionaries!

A small, biology-tuned model designed new protein factors that pushed old cells back to stem cells faster. Meanwhile Apple is testing Gemini as a Siri fallback after internal delays, and Meta is bringing Midjourney’s look straight into feeds and chats.

📌 In today’s Generative AI Newsletter:

OpenAI × Retro Bio AI-engineered SOX2/KLF4 variants hit 50× reprogramming efficiency
Apple explores Gemini inside Siri with privacy guardrails
Meta partners with Midjourney to power AI images across apps

🧬 OpenAI and Retro Bio Rewind the Clock on Human Cells

Image Credit: OpenAI

OpenAI has stepped into longevity science, teaming up with Retro Biosciences to re-engineer proteins that can turn adult cells back into stem cells. Using a custom model called GPT-4b micro, trained entirely on biological data, the project delivered a 50x jump in efficiency over the Nobel Prize–winning Yamanaka factors discovered in 2012. Aged cells showed youthful DNA repair abilities and full stem-cell pluripotency across multiple labs and test methods.

Here’s what changed in the lab:

A new AI model for biology: GPT-4b micro was initialized from a scaled-down GPT-4o and retrained on protein sequences, structural data, and evolutionary context to generate variants with targeted functions.
Re-designed Yamanaka factors: The AI suggested hundreds of new SOX2 and KLF4 sequences. Over 30% outperformed natural proteins, with some variants differing by more than 100 amino acids yet showing stronger reprogramming signals.
Accelerated rejuvenation: The engineered proteins not only triggered stem-cell markers earlier but also repaired DNA damage more effectively than baseline controls, reversing one of aging’s central hallmarks.
Cross-validation: Results held true in human fibroblasts and mesenchymal stromal cells from donors over 50, with derived iPSCs demonstrating stability and the ability to differentiate into all three germ layers.

The collaboration points to a new model for science: instead of massive generic LLMs, small domain-tuned models can solve highly specific problems at speed. For regenerative biology, it points to a near future where the painstaking trial-and-error of protein design gives way to computational search.

Image source: Apple

Apple is in early talks with Google about wiring Gemini into Siri, after its in-house AI push stalled so badly that the next big upgrade is now punted to 2026. Google has already trained a custom Gemini model on Apple’s private servers, with testing underway.

Here’s the play:

Two Siri tracks are running in parallel. Linwood is Apple’s attempt with its own models. Glenwood is the fallback that leans on external partners like Google.
Anthropic and OpenAI are also in the mix. ChatGPT is already quietly powering parts of Siri’s answers.
Apple promises privacy guardrails. Any external model would sit inside its Private Cloud Compute system.
Bloomberg says the call on whether to outsource will come within weeks.

Apple is caught in an awkward spot: Siri desperately needs an overhaul, but the company’s AI bench is thin and rivals keep poaching its best talent. Borrowing Gemini could finally make Siri useful again, or prove how far Apple has fallen behind in the AI race.

Meta just signed a deal with Midjourney to bring its signature “aesthetic technology” into Meta’s AI models and products. Chief AI officer Alexandr Wang framed it as a deep “technical collaboration” rather than a simple licensing deal, hinting that Midjourney’s design DNA will soon bleed across the company’s platforms.

Here’s what we know:

Feeds and chats: Facebook, WhatsApp, and Instagram already test AI image buttons, and now Midjourney will power the look.
Meta’s current suite: Imagine, Movie Gen, and research projects like DINOv3 often lacked Midjourney’s polish.
Midjourney founder David Holz stressed the lab remains “independent, community-backed, with no investors,” even as it partners with one of the world’s biggest tech firms.
Midjourney recently launched its first video model, V1, which can extend images into looping five-second clips.

Meta has poured billions into its AI push, poaching talent and buying half of Scale AI to recruit Wang himself. Now it’s bolting Midjourney’s creativity onto its infrastructure which indicates that even trillion-dollar giants want the indie lab’s magic touch.

GenAI.works is building the infrastructure for the AI agent economy.

The AI agent market could reach $52 billion by 2030*, but most people don’t know which startups will survive. Instead of betting on one, you can invest in the orchestration layer every agent will need.

We have built the world’s largest AI community, with 13 million members and 2,500+ tools in use. Now we are turning that community into the operating system for AI agents:

The GenAI Protocol: an open source engine for agents to collaborate across tools and workflows.
The Distribution Engine: our first application, turning a prompt into a full workflow from finding customers to launching campaigns.

This is not just another AI app. It is the infrastructure powering them all.
Join us as we scale the protocol and distribution engine - Invest today.

Final days to invest. Secure up to 25% bonus shares by August 28. Invest Now.

Reg CF offering via DealMaker Securities. Investing is risky.
*MarketsandMarkets, “AI agents market worth $52.62 billion by 2030,” PR Newswire, July 10, 2024.

Invest Now