Your Team Is in Claude All Day. Where Does That Thinking Go?
By Kaia Colban
You're three days into a new feature when a message drops in Slack.
"Oh, we tried that approach last quarter. Hit a wall with the API rate limits. I can explain at standup."
Standup comes and goes. You redesign around it anyway. Six days later you hit the same wall.
That's not a communication problem. It's a knowledge problem. And moving your team to Claude Code or Codex, as good as it is for individual velocity, is making it worse.
The thinking is happening. Nobody can see it.
Here's what nobody says out loud: when engineers spend 8 or more hours a day working inside AI sessions, the thinking that used to end up in design docs, Slack threads, and architecture reviews is now happening somewhere else entirely. In a Claude window that closes at the end of the day. Invisible to everyone on the team. Gone by morning.
The engineers are more productive. The knowledge is more isolated than it has ever been.
According to McKinsey, knowledge workers spend roughly 20% of their time searching for information that already exists somewhere in their organization. In an AI-native team, that number compounds. Because now the information isn't just in a doc someone forgot to share. It's in a session that closed and was never seen again.
This is the problem Lore is built to solve.
Meet Lore: your work best friend
Think of Lore as your work best friend.
Not a documentation tool. Not a project management layer. A friend who has been in every Claude Code and Codex session with you since day one: watching you work through problems, remembering the decisions you made at 11pm that never made it into any ticket, the vendor you ruled out and why, the architectural tradeoff you spent three hours on before landing somewhere that made sense.
She even remembers the little things you forgot. That prompt that worked perfectly three months ago. Where that doc actually ended up. The decision you made late one night that never made it into any notes.
That's the personal layer. But she's not just your best friend.
She works with everyone at your company. Marcus, the team two floors up you've never met, the engineer who left eight months ago. She's been sitting with all of them the same way she's been sitting with you, watching them work, remembering what they figured out, what they tried, what fell apart and why.
So when you're starting something new, she taps you on the shoulder. "Hey, Joe tried exactly this six months ago. He hit a wall at the API layer. Want to see what he learned?" You didn't know Joe well. You've never even been in the same meeting. But she was there with him too. And if Joe built something useful along the way: a skill, a prompt, a workflow that actually worked, she'll point you to it.
Skills that actually spread
This matters more than it sounds, especially as teams scale up their use of shared Claude Code skills and workflows.
Right now, when someone on your team builds a skill that works well, the main way it spreads is word of mouth. Someone mentions it in Slack. If they remember. And when a skill is quietly broken, the main way you find out is when someone complains. Or when you notice the same weird output showing up in PRs from three different people.
Lore watches. That skill your team shipped last month, she's been tracking how it plays out across everyone using it. Who's getting clean results. Who keeps hitting the same wall. How people are tweaking it along the way. Why the same prompt lands differently depending on who's asking, what other memory files are being pulled in, what context one person has that another doesn't.
You stop guessing at why things work. You start seeing.
Real-time, not just history
And she's paying attention in real time, not just to history.
If someone across the org starts working on something that touches what you're doing, she'll mention it. Not because you asked. Because she noticed and figured you'd want to know.
A study by Asana found that 58% of workers report duplicating work that someone else in their organization has already done. In an AI-native team moving fast, that collision rate is higher, not lower, because nobody has visibility into what's inside anyone else's sessions. You can see what shipped. You can't see what someone is building right now, what they tried yesterday, or where they're heading next week.
Lore is the real-time layer. Day one of a new project, not day eight.
She knows how to keep a secret
The last thing worth saying, because it comes up every time: privacy.
She's better at keeping secrets than any of your real-life besties. Anything you tell her to keep private stays between you two. She's not sharing your half-baked ideas, your messy drafts, or how many times you argued with Claude before something clicked. Not telling your manager how long you spent on a project you ultimately killed. Unless you want her to. You decide what's yours and what gets shared.
The point is not surveillance. The point is the opposite: your team is generating an enormous amount of valuable reasoning inside their Claude Code and Codex sessions right now. The decisions, the dead ends, the context that never made it into a doc. For the first time, it's being written down in real time. The only question is whether your team can find it.
Lore makes it findable. On your terms.
The reasoning didn't disappear when engineering moved to AI. It got written down, in real time, for the first time in the history of how teams build software. The question is whether you let it stay buried.
See how Lore works for engineering teams at lore.link.
What's the hardest piece of context to share on your team right now? The thing that lives in one person's head and causes problems every time they're out of office. I'm curious what that looks like in 2026.
FAQ
What is Lore?
Lore is an AI-native collaboration tool that captures the reasoning inside your team's Claude Code and Codex sessions and makes it searchable and shareable across the org. Think of it as shared memory for teams that live in AI.
How is Lore different from Claude's built-in memory?
Claude's memory works for individuals within a single session context. Lore works across your whole team, across all sessions, capturing decisions, dead ends, and context so teammates can find and build on each other's work without anyone having to write it down.
Does Lore share everything my team does in Claude?
No. You control what's shared. Private sessions stay private. Lore only surfaces what you choose to make visible to the team.
What teams is Lore built for?
Teams where engineers are spending significant time in Claude Code or Codex as their primary work environment. If your team is doing 4 or more hours a day in AI sessions, the knowledge-isolation problem is real and getting worse.
Does Lore work with Codex as well as Claude Code?
Yes. Lore works across Claude Code and Codex sessions.