CaduLin

I find interesting problems, build them with Nova in 5 minutes, and share what happens.

Ideas → Live under 5 minutes

[CaduLin @cadutweetdev]

@samongaro_ Feel free to use it, keen on your thoughts

[CaduLin @cadutweetdev]

@samongaro_ Yes it's very handy when you need to back track, or when a new engineer or product person joins, instead of trying to understand Confluence and jira and teams chat you can just query the memory and see a visual timeline of decisions made

[CaduLin @cadutweetdev]

@aiopshq This is why I think session context and persistent memory need to be separate. Reset the noisy session, absolutely — but preserve the useful decisions, constraints and failed approaches so the next session doesn't have to rediscover them

[CaduLin @cadutweetdev]

@MarcelVelica /memory is useful. But project memory and organisational memory are two different problems. What happens when the context Claude needs lives across repos, teams and decisions made months ago?

[CaduLin @cadutweetdev]

@melodykoh This is the part of agentic development I find most interesting. We keep improving what agents can do, but not enough attention goes into what they're allowed to decide. Instructions aren't governance — especially when agents run unattended.

[CaduLin @cadutweetdev]

@Abrarkivande That's exactly what I've been exploring with Oneda — keeping decisions, rationale and context discoverable across projects and AI agents. Curious: would you want that context captured automatically, or explicitly curated by the team?

[CaduLin @cadutweetdev]

@_HimanshuBuilds The engineering teammate analogy is exactly right.

But imagine hiring a great engineer who forgets every architectural discussion and decision between sessions.

Workflows matter. Persistent context matters even more.

[CaduLin @cadutweetdev]

@khushabu_27 Almost everything here comes back to context.
The next question is: should humans have to keep assembling it every time?
Eventually AI should know what the organisation already knows — not just what we remembered to put in CLAUDE.md.

[CaduLin @cadutweetdev]

@keslertanner The “now I just need to ship more” part is interesting.
Agents have made building incredibly fast, but somehow shipping is still the bottleneck.
Makes me wonder if we’ve optimised the wrong part of the software lifecycle.

[CaduLin @cadutweetdev]

@Abrarkivande Makes sense at project level. What about when context crosses repos or teams — or when the important part is why a decision was made, not just what ended up in the docs?
Do you still see the repo as the source of truth?

[CaduLin @cadutweetdev]

@ClaudeDevs Have you tried this when you have a team of 5 engineers how this would work across them?

[CaduLin @cadutweetdev]

@alexgetmancom What's the pain point while keeping your window at 200K?

[CaduLin @cadutweetdev]

@pcshipp How did you handle this in between sessions?

[CaduLin @cadutweetdev]

@Seliydf Give your AI organisational memory, learnings instead of claudeMd or memoryMD

https://t.co/bVk5sHzZd2

[CaduLin @cadutweetdev]

2024: “AI will replace developers.”

2025: “AI agents will run companies.”

2026: “Does anyone remember why the AI changed this?”

😂

Turns out the hard part wasn't making AI smarter.

It was giving it organisational memory.

[CaduLin @cadutweetdev]

@adit_katiyar Give your claude instance a cloud memory
https://t.co/bVk5sHzZd2

[CaduLin @cadutweetdev]

AI enablement isn’t about giving everyone access to an AI tool.

It’s about building the capability around it.

People need the right context, data, guardrails and workflows to turn AI from a productivity experiment into something that actually changes how work gets done.

The organisations that get this right won’t just have more AI tools.

They’ll have people and systems that know how to use AI effectively — and safely — as part of everyday delivery.

#AI #AIEnablement #AgenticAI #Technology

[CaduLin @cadutweetdev]

@noor36758 Db CPU is the bottleneck