😺OpenAI's new feature (and secret model)

PLUS: Prompt to summarize ANYTHING?!

Welcome, humans.

Happy Friday everyone! Now this is a really fun use of ChatGPT’s image generator—creating fake selfies of yourself with famous historical figures.

What’s fun about this trend is that it doesn’t:

Misuse the likeness of anyone living.
Steal the art style of a famous artist.
Create deep fakes that anyone would mistake for real.

That said, if you try to turn this goofiness into a business (like the guy who made the Ghiblify app), don’t be surprised if the estates of famous dead celebs come for you—if they sue the T-shirt people, they’ll sue you, too.

Here’s what you need to know about AI today:

OpenAI released a new persistent memory feature.
Mira Murati wants $2B for her new startup.
Canva and Airtable launched new AI tools.
Microsoft’s new study put AI debuggers to the test.

Advertise in The Neuron here.

OpenAI expands ChatGPT’s memory and may be testing a new model…

OpenAI is still full of surprises. Yesterday, Sam Altman announced a major upgrade: ChatGPT now has persistent memory across all your conversations.

ChatGPT can now remember details from past chats—even ones from months ago—aiming to create an AI companion that “gets to know you over your life.” The feature is rolling out first to US Pro users (sorry Europe and UK), with Plus users next in line.

Don't worry if this seems creepy; you can opt-out or use temporary chats instead. Though if ChatGPT remembers that time you asked it to write a spicy romance novel about Transformers, maybe it's time for a fresh start anyway.

This announcement came amid buzz fueled by Sam's excited tweets and speculation about new models.

While the soon to launch o4-mini and o3 didn't land today, the real mystery appears to be Quasar Alpha—a “cloaked model” that quietly appeared on OpenRouter earlier this month.

What makes Quasar special?

Massive 1M token context window.
Completely free (during alpha).
Blazingly fast (4x faster than Claude 3.7 Sonnet).
Strong coding abilities (ranking near top models).
Impressive long-context handling.

So this thing popping up to try for free is like the AI equivalent of finding a Ferrari in your garage with an anonymous note saying “test drive me.” I mean, if you insist…

That said, the evidence strongly suggests this is OpenAI's work—the metadata signature, response style, and even Sam's hint (“quasars are very bright things!”) all point that way.

Quasar arrives just as Google is going full Beast Mode on the AI front. We're talking Gemini 2.5 Pro crushing benchmarks, Gemini 2.5 Flash coming soon for speed demons, plus a whole suite of Vertex AI upgrades that would make even the most jaded never-Gemini-er raise an eyebrow.

Microsoft isn't just twiddling its thumbs, either—they've added their own memory features to Copilot, plus vision capabilities and deeper research tools.

Our take: The Memory upgrade is OpenAI doubling down on the “AI bestie” approach—making ChatGPT feel less like a tool and more like a digital companion with a weirdly good memory.

Meanwhile, Quasar Alpha shows they can still drop jaw-dropping tech when they want to (1M tokens and 4x faster? Chef's kiss).

But Google's firing on all cylinders right now—models, hardware, enterprise tools, the works (it even plans to combine its Gemini and Veo models into one).

Is a memory boost and a new stealthy model enough to keep OpenAI in the lead, or are @Sama and company starting to sweat? Plot twist: maybe that's why Sam can't sleep at night.

Read our full analysis of how these releases fit into the AI landscape here.

FROM OUR PARTNERS

Turn AI into Your Income Engine

Ready to transform artificial intelligence from a buzzword into your personal revenue generator? HubSpot’s new "200+ AI-Powered Income Ideas" guide is your gateway to financial innovation in the digital age.

Inside you'll discover:

A curated collection of 200+ profitable opportunities spanning content creation, e-commerce, gaming, and emerging digital markets—each vetted for real-world potential.
Step-by-step implementation guides designed for beginners, making AI accessible regardless of your technical background.
Cutting-edge strategies aligned with current market trends, ensuring your ventures stay ahead of the curve.

Download your guide today and unlock a future where artificial intelligence powers your success. Your next income stream is waiting.

Get Your Guide

Prompt Tip of the Day

Here’s the “ultimate summarizer prompt” we found from this thread. This is a clever way to get your AI to revise its work and iteratively improve its summary outputs.

There are three steps, so make sure you follow them all to get as close to “100% fidelity” with the original source as you can.

P.S: if your source document is really long, have your AI break the summaries up into chunks, and prompt you when to continue.

If it gets confused or loses details in the middle, you may need to prompt it for specific sections with something like this:

“You missed the section on [topic / chapter heading]. Here it is again, pasted below. Please summarize this portion, preserving all… [rest of prompt continues from here]”

FYI: Gemini 2.5 can hone in on a specific section of a long context document if you request it by name, but the other models might need you to paste the missing sections again.

Treats To Try.

Canva AI turns your text prompts into polished designs while learning your style and brand preferences to speed up your creative workflow.
Airtable lets you build and modify apps, create databases, analyze docs, and research web data through chat—free to try.
xAI is finally making its Grok 3 model available via API, giving you direct access to Elon Musk's flagship AI with reasoning capabilities and image analysis.
Artisan’s sales agent Ava finds leads, writes personalized messages, and manages your outreach campaigns (raised $25M).
Voice Notes Pages turns your spoken thoughts into shareable pages where subscribers can listen, read transcripts, and ask AI questions—free to try.
Crono rewrites your sales messages with personalized content from prospect data like job history, social posts, and company news—paid only, starts at €5.
ICYMI: Now that Google’s back to being the GOAT, their 69 page white paper on prompt engineering is probably worth a read.
1. Better yet, copy + paste it into your AI Studio System Instructions, so at anytime you can say “Based on the prompt engineering best practices in your system instructions, write a prompt to do XYZ.”

See our top 51 AI Tools for Business here!

Around the Horn.

That’s an otter lot of progress!

Ex-OpenAI CTO Mira Murati reportedly wants to raise $2B for her startup Thinking Machines Lab at a valuation of $10B.
Google is working with the US’ largest power grid (PJM) to use AI for speeding up grid connections.
Microsoft released a new study where researchers tested AI in a “debug-gym” and found agents with debugging tools showed significant performance improvements over those without—though their simple prompt-based agent still solved fewer than half of SWE-bench Lite issues.

FROM OUR PARTNERS

Get from zero to SOC 2 in 15 hours with Delve AI agents

Delve is an AI compliance platform to help you get compliant with SOC 2, ISO 27001, GDPR and more in 15 hours or less.

Book a demo here for $1000 off compliance and a custom jacket with your logo.

Intelligent Insights

^ that part.

Isomorphic Labs has an ambitious quest to “solve all disease” through AI.
What makes a good leader for people also makes a good leader for AI agents (and vice-versa), according to a new NBER study.
Most tech experts (61%) predict AI will dramatically change human nature by 2035.
New research reveals ChatGPT falls into human thinking traps like overconfidence nearly half the time.
Dan Shipper argues adaptable generalists thrive with AI because it struggles with the truly novel problems where humans excel at connecting diverse ideas.
The IEA warned AI data center electricity use could double by 2030.
U. Illinois researchers used reinforcement learning (AI trial-and-error) to teach language models better multi-step web searching, boosting performance 41%.
Check out Isaac Asimov’s thoughts on AI and how perfectly logical machine brains can potentially “liberate humans and their creativity”—meanwhile, Hacker News thinks we’ve done the opposite.