Catapult handles your most ambitious knowledge work. Currently in Private Alpha.
Early access →
Joined January 2022
- Tweets14
- Following58
- Followers1.3K
- Likes335
What if you could build AI memory systems that thought just like a human..
14 months ago, we set out to build one of the best memory systems in the world.
We believed the context across your work and conversations should compound.
@getCatapult should become more useful every time you use it. It should remember what matters, understand when things change, learn from feedback and forget what is no longer relevant.
At the time, ChatGPT and Claude mostly remembered isolated facts about you. Useful sometimes, but not enough to stop you repeatedly explaining your work and starting each chat from a blank slate.
What are you working towards?
Who owns what?
Why was a decision made?
How have your priorities changed?
What did you correct last time, even if it was subtle?
These details are what make someone genuinely useful to work with.
We went back to first principles. We studied how humans remember and retrieve information, then explored how ideas from retrieval, long-context reasoning and neural memory architectures like Titans and LM2 could translate into an AI system.
Then we tested those ideas against the messiness of real work.
People leave companies. Projects change direction. Documents become outdated. A preference for LinkedIn posts may be completely wrong for an investor email.
Memory is how you filter your past. It cannot be a random collection of facts.
We built systems that could:
→ Separate durable knowledge from what is merely recent, across people, projects, priorities, history and what is currently top of mind.
→ Understand how a small detail changes the bigger picture of a project, relationship or decision.
→ Learn from subtle feedback and corrections, then apply them in the right future situation.
→ Keep memories properly scoped, so guidance for one person, channel or type of work does not leak into another.
→ Track how priorities and active work change, rather than treating yesterday’s focus as permanently important.
→ Build understanding from email, Slack, meetings, documents, attachments, calendars and Notion, not just chat history.
→ Resolve contradictions and replace stale beliefs when stronger evidence appears.
It became an obsession for parts of our team.
Hundreds of experiments. New evaluations built from real work. Graph approaches, hierarchical summaries, ranking systems and attention-based techniques that looked promising but did not survive contact with reality.
We’re now in private alpha. Early users have told us they are “astonished” by what Catapult remembers and how it applies that understanding while helping with their work.
They notice when it remembers how they write, finds a decision everyone else forgot and produces work that is ready to use.
I’d rather show you than tell you.
If you’d like to be considered for our private alpha, let me know.
This screenshot is from one of our first memory prototypes, built over a year ago.
We’re a deliberately small, obsessed team of repeat founders, researchers and engineers from Meta, Amazon Research, G-Research, Oxford, Cambridge and Imperial.
We will transform how the world works. If that excites you, join us!
Most startup careers pages try to convince everyone to join. @getCatapult’s tries to convince you NOT to join us.
We care deeply about building a culture where exceptional people can do the best work of their lives, enjoy doing it and solve problems that genuinely matter.
Culture isn’t a list of values on a wall.
It’s the choices you make about how you work.
The standards you hold.
The behaviours you reward.
And the trade-offs you’re unwilling to hide.
We’re building Catapult with enormous ambition. We tackle problems without clear answers, work closely together in person, take real ownership and care more about finding the truth than protecting our egos.
That environment can be incredibly rewarding. It also won’t suit everyone. So rather than pretending otherwise, we published the reasons not to join Catapult on our careers page.
If the list puts you off, it’s doing its job. If it sounds like the environment you’ve been looking for, we should probably talk.
Check out our new Careers page here: catapult.xyz/careers
Agents trained on the full context of your work are more powerful than anyone realises. Our early users are starting to be blown away by what they're experiencing. Join the waitlist - catapult.xyz
“I'm blown away.”
That message landed two hours after she started using @getCatapult.
Two hours. She hadn’t spent them uploading files, maintaining markdown folders or repeatedly explaining how she works.
Catapult already knew.
It helped tighten her thinking. It caught a duplicate in her schedule that she had missed. It warned her that she had already emailed someone she was about to contact again on LinkedIn.
Small moments. Huge implications. We had hoped people would react like this. We just didn’t expect it this early in our private alpha.
When we started building Catapult, we made a contrarian bet: the biggest limitation of AI agents isn’t intelligence. It’s context.
ChatGPT and Claude are incredibly capable. But most of the time, they start from zero. You explain the backstory. Upload the docs. Paste the thread. Describe the people. Remind them how you like to work. Then do it all again tomorrow.
We didn’t believe a bigger context window or thousands of shallow connectors would solve that. So we chose the harder path. The one most people told us was too difficult.
We built deep integrations across Google Workspace, Slack, Notion, Granola and more. Not simply to search those tools when you ask a question or give it something to do, but to build a secure, living model of your work. One that keeps learning and evolving as you work.
So it already knows your people. Your projects. Your documents. Your decisions. Your tone. Your priorities. Your commitments. The things you said you would do and the things you might otherwise miss.
Then Catapult combines that context with the best frontier intelligence to help produce the work itself. The reply you would have written. The document you didn’t have time to prepare. The missing connection you hadn’t noticed. The next piece of work you need to move forward.
No markdown files to maintain. No skills to write. No rigid folder system. No starting from scratch.
Just an agent that already knows your work.
If you’ve tried ChatGPT Work or Claude Cowork and thought, “This would be incredible if it actually knew me,” that’s exactly why we built Catapult. For ourselves and for people like us.
Judging by the reactions from our private alpha, we’re onto something.
Catapult retweeted
For everyone who’s not CEO of Coinbase or Shopify, soon you'll be able to also have @getCatapult build a model of you from all your work...
Catapult retweeted
The most powerful prompt a founder can ask isn’t “write this for me.”
It’s: “What am I missing?”
I just saw @brian_armstrong describe using an internal AI connected to everything:
Slack, docs, Salesforce, Confluence.
Not a chatbot. A company-aware brain.
Then he asks it questions like:
“As CEO, what should I be paying attention to that I’m not seeing?”
“Where are teams quietly disagreeing?”
“How did I say I’d spend my time… and how did I actually spend it?”
He called it reverse prompting.
Stop telling AI what to do.
Start asking AI what you should be thinking harder about.
That idea made me sit back for a second.
Because most founders don’t need another tool that helps them do more.
They need something that helps them notice.
Notice the patterns.
Notice the blindspots.
Notice the slow-moving risks hiding inside “busy.”
I’ve been using early versions of Catapult for exactly this.
It’s built a model of me + how I work from:
emails, Slack, Notion, docs, and call transcripts.
So I can ask the kinds of questions that are hard to ask humans (or easy to avoid):
“What am I consistently postponing?”
“What decision do I keep circling?”
“What’s the real bottleneck in what we’re shipping?”
And because it actually understands context, it drafts emails in my voice…
Often better than I would…
Without me having to re-explain my world every time.
Private alpha opens in a few weeks.
If you want to try it, comment “CATAPULT” below 👇
Coinbase CEO Explains “Reverse Prompting” and the Rise of the AI CEO
@brian_armstrong:
“One of the big pushes we made in the last year was we got our own internal hosted AI model that was connected to all of our data sources, right?”
“So it's like every Slack message, every Google doc, Salesforce data, Confluence, you know.”
“So now the data is all aggregated and I've started to ask it really… it's not just like prompting it, ‘Hey, can you write this kind of memo for me,’ or something.”
“I'm asking these AI agents now, ‘As CEO, what should I be aware of in the company that I might not be aware of?’ And it'll tell me, ‘Did you know that there's actually disagreement on this team about the strategy?’ And I was like, actually, I didn't know that.”
“This is like reverse prompting. So instead of telling the AI agent what you want it to do, you ask it what you should be thinking more about.”
@Jason:
“It's a mentor. It's a coach.”
Brian:
“Yeah. Like, what could make me a better CEO? And it's like, ‘Well, I looked at how you spent your time in the last quarter and here's how you said that you wanted to spend it, but you actually spent 32% of your time on this instead of 20%.’”
“I've asked it other questions like, ‘What's the thing that I changed my mind on the most over the last year?’ Things like that.”
“It'll prompt you with information you should be thinking about instead of the other way around.”
Thanks to our partner for making this happen!:
Our episode is sponsored by the New York Stock Exchange - a modern marketplace and exchange for building the future. It all happens at the @NYSE.
nyse.com
It's an awesome start to our week at Catapult!
Rundong Liu just started with us as a Founding AI Engineer.
He brings valuable experience in building Cursor-like developer agents on proprietary data, and has conducted independent research in Unsupervised Environment Design. He's also worked at AWS and studied at Oxford and Imperial.
He's already hit the ground running in his first few hours with us! If you want to do your life's best work, building Personalised AI models that understand you, your work and all its context - check out the link in the comments.
#AI #AIEngineer #Startup #hiring #personalizedai
It's such an exciting time at Catapult - we have THREE incredible engineers joining us this month! This week, it's Mayeul, who joins us as a Founding Senior Engineer!
Mayeul brings incredible experience to help build Catapult.
He previously co-founded a deep tech startup where he built a quantum number generator and developed cutting-edge solutions for research teams at Imperial and UCL. He also holds a PhD in particle physics and brings extensive research experience.
If you want to do your life's best work and work alongside Mayeul, building personalised AI models - check out our open roles at catapult.xyz
We're hiring 2 more cracked Founding Engineers!
We’re pushing the frontier of contextual intelligence, by building AI systems that understand you, your work, and its complete context.
We're tackling foundational challenges that Apple Intelligence, Gemini, and others haven't cracked yet.
With Catapult you'll be able to:
▸ Find anything - from specific details to hidden insights in your emails, documents, and apps
▸ Know everything - by leveraging AI across all your data to answer complex questions
▸ Autocomplete work - through smart suggestions and automated actions across your apps and devices
Our early results are promising, we're well-funded, and we're building a team of brilliant minds!
If this sounds like you, or you know the perfect person, check out our roles:
▸ Founding Senior Engineer ➙ catapult.xyz/fse
▸ Founding AI Research Engineer ➙ catapult.xyz/faie
You'll work side by side with @GregWilsenach and @ravsinghsandhu, as we shape the future of Contextual Intelligence!
#hiring #ai #rust #swe #startup #llm
✨ At Catapult, we're hiring Founding Engineers! ✨
We're building AI systems that securely understand you, your work, and its complete context.
We're tackling foundational challenges that Apple Intelligence, Gemini, and others haven't cracked yet. We're pushing the frontier of Contextual Intelligence by combining our powerful multi-stage reasoning models with cutting-edge systems for ingesting and understanding data.
With Catapult you'll be able to:
▸ Find anything - from specific details to hidden insights in your emails, documents, and apps
▸ Know everything - by leveraging AI reasoning across all your data to answer complex questions
▸ Autocomplete work - through smart suggestions and automated actions across your apps and devices
Our early results are promising, we're well-funded, and we're building a team of brilliant minds. We're looking for sharp problem solvers, researchers, and engineers who thrive on tackling novel challenges in a dynamic, highly technical environment.
If this sounds like you, or you know the perfect candidate, check out our roles:
▸ Founding Senior Engineer (Rust) ➙ catapult.xyz/fse
▸ Founding AI Engineer ➙ catapult.xyz/faie
You'll work side by side with @GregWilsenach and @ravsinghsandhu, as we shape the future of Contextual Intelligence!
#hiring #ai #rust #swe #startup
