Another Day, another ridiculous idea. Another win.
This time, at the AINATIVEHACK: SDLC Hackathon, where we asked one simple question:
Why does no one remember WHY we made that decision?
😵 The Problem, Let’s set the scene:
You join a new product team.
You’re looking at some deeply suspicious retry logic in the codebase.
You ask, “Why was this added?”
And then it begins…
Slack messages get forwarded.
Someone linked a Notion doc last edited in 2021.
The engineer who made the decision has moved to Bali and now makes NFTs.
Everyone shrugs and says, “Must’ve been important.”
Welcome to modern software development:
In a world where teams are remotely collaborating, rapidly iterating, and constantly forgetting, we realized something critical was missing:
🧠 Memory.
Not the RAM kind. The “why-did-we-do-this” kind.
✨ The Solution: ContextCache
Enter ContextCache — your AI-powered, passive-aggressive team memory agent.
It silently listens to:
-Slack messages (even those cursed 30-thread debates),
-MS Teams transcriptions (yes, even the ones where Bob forgot to unmute),
-GitHub PRs, Notion docs, and more.
Then it asks itself:
-What was the decision?
-Why was it made?
-Who made it?
-Is this gonna get us sued?
It then summarizes, tags, and stores that info in a vector database. So next time you ask:
“Why did we choose Kubernetes over ECS?”
It answers:
“On April 12, DevOps Dan argued for better autoscaling support in Slack #infra. Legal Lisa also needed HIPAA compliance which ECS lacked. Karen approved the change. You’re welcome.”
🛠️ How We Built It (in 8 Hours Because... Hackathon)
Ingestion Layer
Slack and Teams APIs pumped in conversations and transcripts.
(Yes, some of those were us arguing about what to build.)
LLM Magic via ACI.dev
We piped chunks of conversation into custom agents on ACI.dev, which:
-Extracted decisions,
-Flagged compliance risks,
-Gave structured JSON summaries like a nerdy robot secretary.
Vector Memory Layer
Stored in Pinecone, so we could semantically search decisions by topic, date, or confusion level.
Slackbot Interface
Ask a question like:
“Why did we remove login with Google?”
And boom — instant context.
🥇 The Win
The judges loved it. The devs wanted it.
And the compliance folks? They nearly cried.
We won as An Idea with more Potential
And now we’re thinking — why stop here?
🚀 What’s Next
💬 VSCode plugin so devs can pull decision context without leaving their IDE.
🛡️ Auto-flag compliance issues before audits become horror stories.
🤝 Integrations with Notion, Jira, and whatever comes next.
Our Team: Shivang, Emil, Dave, Mustafa
🎉 Final Thoughts
ContextCache wasn’t just a hackathon project. It was a love letter to every developer who’s ever whispered:
“Why the hell did we do this?”
And now, the answer won’t be:
“¯_(ツ)_/¯"
It’ll be:
“Here’s the full thread, rationale, and approval notes. I got you.”
Top comments (0)