cd /news/large-language-models/how-i-finally-stopped-re-explaining-… · home topics large-language-models article
[ARTICLE · art-34364] src=dev.to ↗ pub= topic=large-language-models verified=true sentiment=↑ positive

How I Finally Stopped Re-Explaining Everything to AI (Bite-size Article)

A developer using Claude's MCP and conversation memory features reports a significant reduction in the effort required to re-explain context, leading to faster and deeper conversations. By connecting Notion and Google Docs in read-only mode, the AI can reference documented project details across sessions, eliminating the need to re-introduce context each time. The developer notes that while the features require active management to avoid over-connection or stale references, the overall shift has made interactions more efficient and natural.

read4 min views2 publishedJun 19, 2026

As I mentioned in my last post, I've recently started making more serious use of MCP and conversation memory in Claude. Neither of these are new features — they've been around for a while. I just finally got around to actually using them.

The changes turned out to be subtle but significant. My relationship with AI, or the sense of distance between us, shifted more than I expected.

At the core of it, I think it comes down to the elimination of explanation cost — the effort of having to re-explain everything from scratch every single time.

I wrote about MCP in my previous article, so consider this a continuation of that.

(MCP is roughly a way to connect Claude to external services. Conversation memory lets it carry context from previous sessions into new ones. That's enough background to follow along.)

For a long time, my approach was simple: just copy and paste whatever context I needed into each session. I was curious about MCP, but honestly, I didn't think the lack of context continuity was that big a deal. Yes, copying things over was a bit of a hassle — but you get used to it. It was just part of the workflow. Then I actually started using both features, and I noticed things I hadn't before.

Looking back, it was tedious work. The kind you don't really notice until it's gone.

Now I have Notion and Google Docs connected in read-only mode. As long as I've documented a project's direction, naming rules, and key decisions, Claude can reference that in any session — and the context stays consistent.

There's a prerequisite to this working, though. I've always had a habit of writing things down, even for solo projects. The idea of plans drifting or decisions being forgotten bothers me, so I keep my own documentation in Notion as a matter of course. That habit turned out to be the foundation that made MCP actually useful. Connecting MCP is only half of it — it needs something worth referencing on the other end.

Conversation memory has also had more impact than I expected. When I ask "didn't we talk about this before?", it usually remembers accurately. And sometimes it picks up on context without me even asking — it just knows, and the conversation moves faster because of it.

Things I used to explain every time are now things I don't have to explain at all.

When two people talk, the conversation carries forward. You don't start from scratch every time you meet. But that's exactly what AI used to require. Every new session, the previous one had never happened. Same personality, same tone — but no memory of you. You had to re-introduce yourself every time.

That gap was particularly noticeable because the in-session experience felt so natural. Mid-conversation, it really did feel like talking to someone. But the moment a session ended, that sense of continuity broke.

With MCP and conversation memory, that gap has largely closed. The starting point of a conversation has changed. Before, I'd spend the first part of any session getting Claude up to speed on who I am and what I'm working on. Now I can start from a place of shared understanding — and go deeper, faster.

Since I started using both features, I reach useful answers more quickly, and there's less back-and-forth to get there. The density of each conversation has gone up.

That said, it takes some active management. Conversation memory can sometimes over-connect things — drawing links to past discussions that aren't really relevant. And the documents I've connected via MCP can go stale if my thinking has evolved. With both, I've found it's important to stay aware of whether what Claude is referencing still reflects where I actually am. This isn't a passive tool. You have to know where you want the conversation to go.

One more thing: I still haven't given Claude write access to anything. Notion and Google Docs are both read-only. And yet the difference in experience is real. I don't feel any urgency to change that — for now, this is more than enough.

I'm still early in using these features, so I don't think I've found the final answer yet. I'll keep at it, and write again if anything new comes up.

Thanks for reading :)

── more in #large-language-models 4 stories · sorted by recency
── more on @claude 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/how-i-finally-stoppe…] indexed:0 read:4min 2026-06-19 ·