Launching today: Instructions vs. Memory

Context v2 for Pretty Prompt 🧠.

Not all context is equal. So we stopped treating it that way.

Every power user of AI eventually hits the same wall.

Everything sounds sh*t unless the AI knows you.

“Knows you” means it needs context. It needs rules. It needs memory. It needs examples. The LLM needs some kind of guidance or parameter of what good looks like, for you.

This is where Context comes into play.

But most tools treat all context the same. A rule and a memory carry equal weight. A formatting preference and a career background are stored identically.

We think that’s wrong. Or at least we learnt this the hard way 😅.

And we just shipped something to fix it.

Meet Context v2.

---

Introducing: Instructions vs. Memory

Today we’re launching what we’re call Context v2 (maybe we’ll nail it this second time) — a smarter way to classify the context you add to Pretty Prompt. And for it to make a real difference into your AI outputs.

Here’s the split:

🔒 Instructions: explicit rules.

Applied exactly as you wrote them, every time.

• “No em-dashes.”

• “Always respond in under 512 characters.”

• “Use bullet points.”

These are your non-negotiables. The rules you’d give a new hire on day one. They don’t bend based on context, they are explicit. Applied as you wrote them.

🧠 Memory: implicit context.

Woven in only when it’s relevant, as part of the prompt.

• “I’m a solo founder building a B2B SaaS for law firms.”

• “My audience is mostly non-technical.”

• “I’m working on a client proposal that includes the following sections…”

This is background. It shapes your outputs without appearing in them. It’s the difference between a tool that knows about you and a tool that actually knows you.

---

The part we’re most proud of: you don’t configure this.

You just add context. We auto-classify on the backend.

When you write “I like my answers in bullet points, and in under 512 characters” — the tool reads that as a rule. It files it as an Instruction.

When you write “I lead product at a 20-person fintech startup called Fintitech”, the tool reads that as context. It files it under Memory. It’ll surface when you’re writing a product brief. It’ll stay quiet when you’re planning a holiday trip.

This is what makes your slides different from mine. Your marketing plan different from someone else’s. The context layer is where personalisation actually lives, not in the model, but in what the model knows about you.

I can tell you this is the same for every AI tool on the internet. The more context it has about you, your team, or your company, the better responses it will give.

Try it out on the Web App/Context. (Syncs with your extension by itself.)

---

Why we built this

Pretty Prompt has always been about making your prompts smarter. But we realised, better prompts are just one piece of the puzzle. The real “magic” stays with that context. Now you can take that context with you, regardless of the AI you use.

It’s a small update in terms of code. It’s a big update in terms of how the tool thinks.

Still bootstrapped. Still shipping. 🚀

---

If you haven’t tried Pretty Prompt yet, give it a go; it’s free to start, and we’ve made Pro free for 24 hours for new users. No excuses ;)

👉 pretty-prompt.com

Thank you for building this with us. 💜

Ilai

(PS: I share daily updates on LinkedIn, give us a follow if you want the compressed version 🫰)

---

Review of the week:

“Excellent software which really improves my workflow. The quality of the prompts it returns is simply amazing.

Definitely a 5 star product and highly recommended. The support and speed of the response from the team is excellent”

---