In partnership with
In Today’s Issue:
🚀 OpenAI is back: GPT-5.2 delivers a major jump in reasoning and reliability
🤝 Gemini 3 Pro earns the highest-ever public trust score
🛰️ SpaceX is reportedly preparing for a record-breaking $1–1.5 trillion IPO
✅ FACTS Benchmark reveals that all leading models still struggle to reach 70% accuracy
✨ And more AI goodness…
Dear Readers,
What if today is the day “AI assistant” quietly becomes “AI colleague”? In this issue, we dive into GPT-5.2 and what it really means when a model can run projects end-to-end instead of just answering your next prompt. Around it, we zoom out to the wider landscape: Gemini 3 Pro earning record-high trust scores in a massive real-world study, SpaceX lining up what could become the biggest IPO in history, and DeepMind’s new FACTS benchmark exposing where today’s models still hallucinate more than we’d like. To round it off, you’ll get a robotics-lab tour with Hannah Fry, a fresh prompt of the day, and a dose of rumors and leaks from the AI trenches. If you want to understand not just what is happening in AI but how it will touch your work and future, this is your read… Let’s jump in.
All the best,
Kim Isenberg
Gemini 3 Pro Earns Trust
Gemini 3 Pro achieved a 69% trust score in a massive blinded study with 26,000 participants, up from just 16% for Gemini 2.5 Pro—making it the most trusted model ever measured by Prolific’s HUMAINE framework.
The model now ranks #1 across demographics for perceived trust, ethics, and safety, suggesting that Google’s focus on real-world user evaluations, not just academic benchmarks, is paying off and could reshape how enterprises choose AI systems.
SpaceX Targets Record-Breaking Mega IPO
SpaceX is finally preparing to go public after years of avoiding the stock market, with reports pointing to a 2026 IPO aiming to raise more than $25–30 billion at a valuation around $1–1.5 trillion, potentially the largest IPO in history, eclipsing even Saudi Aramco. The fresh capital would supercharge Starlink’s global internet rollout, fund Starship and even space-based data centers, while giving employees and early investors real liquidity after years of private growth, as SpaceX’s revenue is projected to climb from about $15 billion in 2025 to over $22–24 billion by 2026. Now even confirmed by Musk himself.
FACTS Benchmark Reveals Truth Gaps
Google DeepMind’s new FACTS Benchmark Suite evaluates how factually accurate large language models are across four dimensions: parametric knowledge (what the model “knows” internally), search-augmented answers, multimodal (image+text) questions, and grounding on long documents, using 3,513 carefully curated examples plus a hidden private test set. Gemini 3 Pro tops the initial leaderboard with a FACTS Score of 68.8%, cutting error rates vs. Gemini 2.5 Pro by 55% on search tasks and 35% on parametric questions, but all leading models still stay below 70% accuracy, showing that hallucinations remain a hard unsolved problem.
Google DeepMind robotics lab tour with Hannah Fry
OpenAI is back - GPT-5.2 is a big release!
The Takeaway
👉 GPT-5.2 delivers a major jump in reasoning and reliability, making AI agents viable for real professional workflows rather than short prompt sessions.
👉 long-context performance (256k) and fewer factual errors mean teams can centralize documents, decisions, and research inside one persistent agent.
👉 Benchmarks like GDPval, ARC-AGI-2, and SWE-Bench verify that GPT-5.2 can execute multi-step knowledge work with higher accuracy and lower oversight.
👉 Companies can now automate entire workflows—modeling, coding, slide creation—turning GPT-5.2 from a helper into a productive, ROI-positive digital teammate.
GPT-5.2 is the point where “chatbot” quietly turns into “colleague.” OpenAI’s new flagship is tuned for serious professional work and long-running agents: it builds banking models, refactors codebases, assembles polished slide decks, and keeps track of huge, messy contexts instead of dropping the thread after a few prompts. In practice, it feels less like a Q&A box and more like a junior teammate who can sit inside your workflows all day long.
On the numbers, GPT-5.2 Thinking sets a new bar across economic and reasoning benchmarks. It beats or ties human experts on about 70.9% of GDPval knowledge-work tasks, reaches 55.6% on SWE-Bench Pro and 80% on SWE-Bench Verified, hits 100% on AIME 2025, and more than triples GPT-5 on ARC-AGI-2—the exact evals people watch for “can this actually do my job-like tasks?” At the same time, OpenAI reports roughly 30% fewer factual errors and near-perfect retrieval across 256k-token contexts, making long-document workflows actually usable instead of fragile experiments. GPT-5.2 Instant, Thinking, and Pro start rolling out in ChatGPT for paid plans today and are live in the API for builders.
GPT-5.2 is a big surprise and a big jump for OpenAI
Why it matters: GPT-5.2 doesn’t just answer questions; it runs projects end-to-end, from spreadsheet to slide deck to code. For builders, teams, and solo creators, that’s the shift from “AI as a clever assistant” to “AI as an always-on teammate with real economic impact.
Turn AI Into Extra Income
You don’t need to be a coder to make AI work for you. Subscribe to Mindstream and get 200+ proven ideas showing how real people are using ChatGPT, Midjourney, and other tools to earn on the side.
From small wins to full-on ventures, this guide helps you turn AI skills into real results, without the overwhelm.
Can Rapamycin Reverse Hair Loss?
For years, rapamycin has been the quiet star of longevity circles, now it’s crashing the hair-loss world. A growing wave of studies suggests that dialing down the mTOR pathway with rapamycin can push hair follicles back into growth mode, reduce “aged” cells in the scalp, and even restore pigment in graying hair. Ex vivo work on human hair follicles shows that inhibiting mTORC1 can prolong the anagen (growth) phase and boost pigmentation, hinting that the same molecule used to slow cellular aging might also reboot tired follicles.
Researchers are getting creative with delivery: dissolvable microneedles that co-deliver rapamycin and EGCG, rapamycin-loaded microdepots injected into the skin, and exosomes “primed” with rapamycin to supercharge regrowth in animal models. Longevity-focused clinics are already experimenting with high-strength topical formulas that combine rapamycin with antioxidants to reduce follicle senescence and inflammation in humans.
It’s still early days – dosing, safety, and long-term effects need hard clinical data – but the idea that a classic immune drug could become the backbone of next-gen hair therapies suddenly feels less like sci-fi and more like roadmap
Sources:
🔗 https://www.gethealthspan.com/research/article/rapamycin-for-hair-loss-research
🔗https://pmc.ncbi.nlm.nih.gov/articles/PMC10328083
🔗https://pmc.ncbi.nlm.nih.gov/articles/PMC9318735
Join the 1% who won’t sleep on the future.
Mindstream is the human-first AI newsletter that prioritizes clarity over complexity and progress over perfection—so you can make smarter moves, lead calmer conversations, and keep your team a step ahead, today.
