I love memory because it points to where we will hopefully go with the product. You will have this entity that knows you and connects to all your stuff — Sam Altman

In a recent YC interview, Sam Altman mentioned something that felt small at first. Just one sentence, “I love memory because it points to where we will hopefully go with the product. You will have this entity that knows you and connects to all your stuff”. The conversation moved on but the idea stuck with me. ChatGPT is already your identity. What if you were using it as “Sign in with OpenAI” button with other applications?

If you’re building a product right now, pause. Ask yourself:

• Why does my product’s onboarding exist?

• Why does my product ask the user who they are, what they want, and what they know?

• How will the product change if I already knew all that?

That’s what one-click “Sign in with OpenAI” offers. The user arrives with context. Past decisions. Preferences. Writing style. Task patterns. Maybe even mood.

You’ll be building interfaces that assume the user starts mid-conversation.

It’s not just about personalization. Most SaaS apps today assume a blank user.

But, what happens when the user shows up already loaded?

It won’t just be a login button. The implications are much more, bending everything the software assumes about you.

It plays a little with interoperability in the OS, but it’s different. Browsers are more aligned with OS concept, like Dia or Action browser, but individual products can also leverage the user’s deep context.

When you sign into most products today, you bring an email address and a trail of past preferences. Maybe the app remembers your theme, your last session (like Lovable’s “last used”), or your keyboard shortcuts. But that kind of personalization is shallow. It’s a scrapbook of clicks and toggles.

Lovable shows what you used to last login — source author

ChatGPT already holds something different. It remembers how you explain problems. How formal or chaotic your writing voice is. Whether you like being challenged or reassured. For many users, it holds emotional residue. Sessions with the LLM have started to carry the feel of internal dialogue. Over time, that residue solidifies into a cognitive pattern. When you sign into a product through OpenAI, that pattern could come with you.

Your thought patterns, your values, your obsessions, your humor, your blind spots.

That changes what “Sign in” even means.

Why do I have to tell each app I want a 2-bed in Tokyo?

Righ now:
I search for a place to stay in Tokyo. I pick the filters
- 2 bedrooms
- Entire home
- Quiet area
- $100–150/night
- Must have Wi-Fi + washing machine

I apply the filters, skim photos, reject based on lighting or layout.

Then, I check prices on Agoda. I re-enter everything.
Again on Booking.com.
Again on Expedia

With “Sign in with OpenAI”:
It knows my preferences from past travel
- “always wants natural light”,
- “walks everywhere”,
- “hates hotel lobbies”

It ports the intent and shows me stays according my liking.

Pre-applied. Already loaded. I just have to book.

It even adjusts filters based on context: “You’re traveling with friends this time, so I expanded the budget slightly and prioritized shared living space.”

🤯 A crazy thought

We’re all making first drafts right now with a prompt. Using “Sign in with OpenAI”, you may even have the first draft waiting for you!

The consumer layer

Each generation using AI represented graphically — source author

Tools learn you instead of just serving you.

OpenAI has already become a kind of cognitive gym for millions of users. Especially the new generation. Sam Altman has talked about the generational differences he’s observed.

• Gen Z treats GPT like a personality mirror, a digital twin. They train it to sound like them, argue like them, reflect their aesthetic and emotional register.

• Millennials use for work and making important life decisions. Relationship advice? How do I tell my boss I’m leaving? Should I tell my friend I saw him stealing?

• Wiser users, always practical, treat GPT like a turbo intern who also happens to write copy and understand burnout. A cheat code for work and productivity.

• And our beloved parent generation? They’re finally using a computer assistant without needing to call you first. Progress.

If this model becomes portable, move from ChatGPT into other products, then “signing in” starts to feel more like bringing your personal OS.

Imagine if every app knew you every time you opened — source author

Imagine…

…if opening Spotify and it plays music that matches your current mood, inferred from recent ChatGPT sessions.

…if Duolingo toned it down a bit today because it picked up on your quiet, tired tone. You never said anything. But it knew.

Every app becomes a context-aware extension of you.

Products stop offering generic interfaces and begin surfacing pathways that reflect your current cognitive state.

If “Sign in with OpenAI” could exist in 12 months

then we need to start designing as if it already does.

You now move from identity-as-auth to identity-as-a-pattern.

1. Intent-first product thinking

• What state of mind is the user likely in when they show up here? (e.g., anxious? goal-driven? exploring?)

• What recent decisions have they made elsewhere (e.g., in ChatGPT or another tool) that should influence this experience?

• How would a trusted assistant describe why the user opened this product today?

2. Start mocking interfaces that skip onboarding

• There’s no tutorial. The assistant teaches by doing.

• There’s no first-run checklist. The assistant preloads tasks based on your last 5 prompts.

• The product opens in the middle of something the user already started — even if that “start” happened in ChatGPT, not your app.

3. Let the assistant be the identity

Design features where the assistant is the carrier of truth, not your backend:

• The assistant remembers what the user does (not you).

• The assistant stores goals, drafts, projects, unfinished plans and your app just renders them when summoned.

Does that mean you feed info back to the identity? Maybe yes. Make the identity assistant even more powerful. (Am I being anti-capitalistic?)

How do you get started today?

Don’t worry, I got you :)

A glimpse of the cheatsheet — source aiverse

I created a “8 experiments to run today” cheatsheet for you to start building for the future.

👉 Grab the free cheatsheet

The enterprise layer

Context transfer, strategic memory, delegation.

In business software, the shift takes a different form. The core problem in most enterprise tools is keeping everyone updated. Context is scattered across Slack threads, Notion pages, ticketing systems, and meetings. Employees repeat themselves constantly.

Now consider what OpenAI has (or another dominant that may emerge in the business land). People feed GPTs the content of their jobs. The assistant has seen the quarterly roadmap, your last review meeting, the next action steps, the upcoming meetings, the timeline, the hiring plan, or the investor update. It contains the meta-layer: how you reason through the information.

“what if” conceptualization — source author

🤔 What if a team member can bring their GPT-trained model into a tool, then that tool gains access to how they think. Not just what they wrote last week, but how they tend to respond under pressure. What tasks they procrastinate on. What signals they trust.

Powerful but a lil scary too! With great power, comes great responsibility.

Documentation tools that write in the person’s voice, while remembering which details they tend to gloss over.

🤔 What if, hear me out, you ‘Sign in with OpenAI’ on a platform that checks for culture fit?

Company-wide GPTs could emerge as a shared framework. They could absorb onboarding content, internal playbooks, tone guidelines, and meeting summaries. When new hires log in, they receive guidance that reflects the actual behavior of the team, not the sanitized documentation version.

There’s a big hefty market here, and a play that’s already afoot.

Salesforce recently made headlines by blocking other AI vendors from accessing Slack data. The reason is obvious. Slack is a firehose of raw organizational thinking. It captures unfiltered work energy: ideas in motion, pressure in real time, negotiation before it’s formalized. Losing access to that data means losing the edge in AI tooling.

So Salesforce did what any self-protective platform would do. It closed the gates. This positions them them directly against OpenAI but with a huge headstart.

For me, as I too build my own AI Designer, an important question I keep asking myself is, while big companies fight against OpenAI, what should new startups do?

Should you fight to own user’s identity?

You always build for 3 years down the line. And in order to not get replaced by an LLM’s latest version, one should build with the hypothesis that the LLM will get 10x better. Will your product get a boost or will it be replaced?

🤔 What if you just accepted OpenAI’s long play?

A good example, and my favourite AI product, is Granola. They didn’t solve for their small input context length for 6 months! They could only transcribe meetings under 30 minutes and the MOST requested feature was larger context (longer videos). They held their ground until OpenAI released a new version and they instantly upgraded. Why build against OpenAI when you can leverage it?

Opting out of the race

Every product wants to become the center of gravity — source author

Most platforms are still scrambling around to own the user. The modern SaaS playbook involves capturing data, learning preferences, locking in workflows, and quietly reinforcing dependency. This is why products ask for logins that go through their own authentication gates. If they can be the entry point, they can be the platform. That’s the moat in today’s era of AI-driven products.

But what if tools stopped fighting for owning user’s context and just plugged into the one the user already brings?

Imagine a future where software stops building isolated memory systems and instead aligns itself to an “identity container” provided by OpenAI. Every product would start with the assumption that the assistant understands the user better than the product ever could.

Sounds utopian? It’s also deeply contradictory to how most companies operate today.

Every product wants to become the center of gravity. Every platform wants to be the “source of truth.” But if “Sign in with OpenAI” becomes the dominant way people bring themselves to software, then these data silos start to look obsolete. Instead of competing to understand the user, apps can simply align to the assistant that already does.

From personalization to multiplicity

One cognitive model won’t be enough. People move through roles. You’re a strategist in one app, a creative in another, a supportive manager in a third. Logging in with OpenAI doesn’t just require bringing yourself. It also requires choosing which version of yourself to bring.

Instead of one login, people begin maintaining profiles — source author

The fragmented self

Most people already operate across multiple identity layers. There’s the personal Google account for spam, the work Google account for meetings, the burner Gmail used for shopping. Each one holds a piece of the user, but none hold the whole. People switch between them constantly. They drag files between drives. They lose bookmarks in the wrong Chrome profile.

Systems force users to split, so they adapt by compartmentalizing. But these are just interface divisions.

This is where modular identity starts to form. Instead of logging in with one fixed assistant, people might begin maintaining profiles tuned for different cognitive and emotional states. You sign into your productivity suite with “Focus Mode,” a sparse, silent assistant that prioritizes and filters without commentary. Later, you drop into a collaborative doc using “Mentor Mode,” which surfaces analogies, alternative perspectives, and coaching language. A completely different version of you shows up.

That version might carry different permissions, memory scopes, or vocabularies. Some might even start out as templates: users adopting pre-trained personality overlays for specific roles, like “head of remote ops” or “early-stage founder under pressure.” These become shareable across teams. Identity becomes a tool.

The new login

What OpenAI potentially offers is a chance to unify these fragments under a portable identity. You could bring the same assistant across domains, but filter its behavior through modular personas. One GPT knows how you behave in meetings. Another carries your social tone. A third exists purely for internal reflection. They all originate from the same identity but behave as distinct agents based on the settings.

Logging into software would become more like shifting gears — source author

Instead of managing multiple logins, users would begin managing versions of self. Logging into software would become more like shifting gears. The identity would travel with the user, not tied to an app.

From GUI to Ambient UI

This is where it gets real

Right now, all of this lives inside a UI. But OpenAI and Jony Ive recently announced their merger to build a personal device (just speculation). Nobody knows what it looks like yet. But Jony Ive designs objects that disappear into behavior.

OpenAI’s $6 billion collaboration with Jony Ive surely isn’t about making a better iPhone. My hunch: it’s about building a shell around this “identity”. If “Sign in with OpenAI” is the access layer, this hardware could become the memory intake valve. A way to inject reality into your GPT.

While we don’t know what the device is, it is coming — source author

Because right now, your assistant only knows what you type into it. But most of your life happens outside the text box.

• The tone you used in a meeting.

• The moments you chose not to respond.

• The confused look when someone over-explained something simple.

The parts of the day you avoid thinking about.

A Ive-designed object could start to capture those things. Not in a dystopian, surveillance-state sense, but in a consensual, intimacy-focused way. Ive designs touchpoints; receiving data from gesture, voice, silence, movement, and mood. This data becomes part of your identity.

It brings your assistant to the physical world and allows it to feel you more fully. In case you’re thinking what I’m thinking, then yes, Her 2.0.

Identity as the future

At this point, the question for every software company becomes simple. Will you build your own isolated user identity, or will you accept the GPT model as the new standard?

Salesforce is choosing the former. It’s blocking access to Slack data, preserving its moat. And it makes sense. They have the talent, the resources, and the platform-era dominance; whoever owns the data owns the intelligence, hence, owns the user.

But if the user owns their own data (even though via OpenAI), then the game stops being about hoarding and starts being about welcoming. Instead of guarding your castle, you design your front door to accept the wandering minds who arrive with their own models.

Products that get this early may start looking lightweight, almost empty. But that emptiness is intentional. It makes room for the user to bring themselves in full fidelity.

Interfaces that ask “Who are you?” will now ask “Which version of you is here?”

The product listens first. Then it builds itself around the answer.

And all it might take is a ‘Sign in with OpenAI’ button.

The magical button coming soon — source author