Drawsy AI

An AI workspace where your canvas, context, code, and live software work together.

[Drawsy AI @DrawsyAI]

We're open-source btw!

[Drawsy AI @DrawsyAI]

Kudos to the amazing team at HydraDB @contextkingceo @iHarnoorSingh @abhirupvg

[Drawsy AI @DrawsyAI]

4/7
Personal memory is private and tied to the signed-in user.

Connector knowledge is external project context.
They work together, but they don’t become one blurry "AI memory."

[Drawsy AI @DrawsyAI]

Built by HydraDB for #HackHydra and later in preview to all at https://t.co/RpNvSHUm17

[Drawsy AI @DrawsyAI]

3/7
When a question needs current facts, Drawsy can retrieve context from connected Gmail, Notion, GitHub, Calendar, Drive, and more.

It shows the source instead of making a confident guess.

[Drawsy AI @DrawsyAI]

1/7
The thing we hate about AI assistants isn’t that they forget everything. It’s that they forget the one decision that mattered right when we move to a different canvas.

That’s why we built Drawsy.

[Drawsy AI @DrawsyAI]

2/7
Drawsy lets us make a decision in one canvas, open another, and continue without copy-pasting a transcript or using a magic command.

The context follows the work.

[Drawsy AI @DrawsyAI]

@spakhm There’s a lot of nuance here. We tried turning the whole thing into a visual system the context, sequence, and failure modes are much easier to see this way made with https://t.co/RpNvSHUm17

[Drawsy AI @DrawsyAI]

@Awesome_O_AI 100%.

A system design answer is easy to judge on the final diagram.

The real test is whether the AI can explain why each piece is there and change it when the requirements change.

[Drawsy AI @DrawsyAI]

We gave Drawsy one system design question:

“Design a news feed system.”
Requirements:
- mobile app as well as web app
- user can publish a post and see friends posts on news feed page
- feed is sorted by reverse chronological order
- User has 5000 friends
- traffic volume is 10 million DAU
- feed can contain media files including both images and videos Can you propose a High-Level Design (HLD) for this news feed system?

It gave us this.

API → feed service → event bus → fanout → Redis → DB → read path → media pipeline.

The fun part starts now:

Can we challenge the architecture, change it, and keep the reasoning on the same canvas?

https://t.co/tlOUqpaqq7

[Drawsy AI @DrawsyAI]

@justinskycak This is why good tools matter more than good intentions.
If the system keeps the right context visible and the next action obvious, people don't have to be heroic just to do good work.

[Drawsy AI @DrawsyAI]

The next AI war won't be model vs model.

It'll be context vs context.
Your codebase.
Your decisions.
Your docs.
Your GitHub.
Your Slack.

What changed yesterday.
Why it changed.

Give two agents the same model but different context and you can get two completely different outcomes.

The model is only half the product.

[Drawsy AI @DrawsyAI]

@rohanpaul_ai If software becomes abundant and opensource, the interesting question becomes: what becomes scarce?

Compute, energy, distribution, and context all start looking a lot more important.

[Drawsy AI @DrawsyAI]

@agazdecki The weird thing about startups is that the failed ideas often become part of the context that makes the next one work.
You don't really start from zero.

[Drawsy AI @DrawsyAI]

Feels like the next phase of AI coding is becoming pretty clear.

The model isn't the whole product anymore.

Give an agent a repo and it'll write code.

Give it the repo + decisions + issues + docs + history + the reason things are the way they are...

now you're giving it a chance to actually understand the software.

That's a very different thing.

[Drawsy AI @DrawsyAI]

@Vatsalpandya333 this is the shift I'm most interested in.

the useful AI engineer won't just answer "how do I fix this?"

it'll be able to trace the question across the actual system and figure out what changed, where it broke, and why.

less chatbot. more investigator.

[Drawsy AI @DrawsyAI]

@hnasr I usually want to find the hard part early.

Not necessarily solve it completely, but understand it well enough to know whether the rest of the system is actually going to work.

Shipping the easy 80% feels great until you discover the hard 20% changes the whole architecture.

[Drawsy AI @DrawsyAI]

this is exactly the problem we're thinking about with Drawsy.

the useful context for an AI coding agent isn't just the prompt.

it's your GitHub repo, architecture, code, diagrams, decisions, and how everything connects.

we're building Drawsy to bring that context together so AI can reason about the system, not just generate another snippet.

https://t.co/pQLdi5nbNo