Every engineer hits a point where they become the bottleneck.
Mine came on a Friday afternoon — the exact moment my brain had already checked out and checked into the weekend. I had three Slack messages waiting, a half-written bug fix open in my editor, a Jira ticket someone had been chasing for two days, and a pull request sitting unreviewed since Wednesday morning. Our team was scaling a security platform composed of 20+ microservices, with a SOC team using and testing it, and somehow I was expected to be in the code, in Jira, in GitHub, and in Slack — simultaneously.
I wasn't dropping the ball. I was juggling too many balls to catch any of them well. So I did what any engineer would do when a problem is bad enough. I built a solution.
Meet E.D.I.T.H 🤖
(hahaha ik credits to marvel)
E.D.I.T.H is an AI that lives in my Slack workspace. It creates Jira tickets, reviews pull requests, investigates and fixes bugs, sets deadlines, answers questions about the codebase, and keeps my team moving — without me being in the room.
I didn't build a chatbot. I built a representation of myself — powered by Claude, Anthropic's AI — one that knows my systems, speaks in my voice, and does what I would do if I were available every minute of the day.
E.D.I.T.H works through a set of skills. Each skill is a specific, focused capability — like fixing a bug or reviewing a PR — that my team or I can trigger directly from Slack. No dashboards. No portals. Just a message, and E.D.I.T.H handles the rest.
Here's what makes it actually work: every skill is something I used to do myself. When I caught myself handling the same type of request repeatedly — triaging a bug report, reviewing a diff, setting a deadline and notifying someone — I wrote down exactly how I would do it. Not code. Just instructions: what to ask, what to check, what to do with the answer. Claude runs those instructions.
The output is what I would have done. That's the whole idea. E.D.I.T.H's capabilities aren't general AI capabilities. They're mine, written down and handed off. The team still depends on me — they always will. But now that dependency doesn't require me to stop what I'm doing to show up.
Every action E.D.I.T.H takes sends me a DM. Every ticket, every PR, every deadline update — there's a notification trail that leads to me. E.D.I.T.H doesn't make decisions. I do. E.D.I.T.H carries them out.
But Why????????? 🤔
The Interruption Tax Nobody Talks About
Research puts the recovery time from a single interruption at around 20 minutes. Not the interruption itself — the getting back to where you were. Multiply that by ten Slack messages, three ticket requests, and two "quick questions" and you've quietly lost half your day before you've shipped anything meaningful.
The problem wasn't that my team needed things. The problem was that I was the only path to getting them.
E.D.I.T.H changes that equation. My team still depends on me — on my judgment, my standards, my approval. That hasn't changed. What changed is that they no longer need me to be present and available to get what they need. My thinking is already in the skill. E.D.I.T.H is just the part of me that's always there when I'm not.
Skills That Changed My Day 😊
!bug — From Report to Pull Request, Automatically
This is the skill that surprised me most — not because of what it does, but because of how much it does.
When my SOC analyst finds a bug in our product, they don't ping me anymore.
They type !bug and describe what they saw. E.D.I.T.H asks a focused question or two — which customer is affected, what the expected behaviour was — then creates a properly structured Jira ticket: typed, prioritised, assigned, added to the active sprint, and given a deadline based on severity. High priority? Due within the hour. Medium? End of day.
But it doesn't stop there.
E.D.I.T.H then investigates the bug in the codebase. It traces the failure, identifies the root cause, writes a fix, and commits it to a feature branch. It reviews its own code before raising a pull request — checking for correctness, edge cases, and security implications. Once satisfied, it sends me the PR.
What used to take me an hour of context switching — read the report, reproduce it, find the root cause, write the fix, review it, raise the PR — now lands in my Slack DM as a ready-to-review pull request. I look it over, approve it, and move on.
The analyst gets a Jira confirmation. I get the PR. Everything in between was E.D.I.T.H.
!review — Nothing Merges Until It's Actually Ready
Submitting code for review used to mean waiting on me. Now it means triggering a skill that doesn't stop until the code is genuinely clean.
When someone raises a pull request, the !review skill reads the entire diff carefully — checking for bugs, security vulnerabilities, logic errors, and whether the implementation fits the patterns of the existing codebase. If issues are found, they're posted as review comments directly to GitHub.
The developer fixes them and pushes again. Then E.D.I.T.H reviews again.
This loop — review, fix, commit, second review, safe to merge — runs until the code passes.
Only then does E.D.I.T.H report back that the PR is ready. Nothing reaches production having had only one set of eyes on it. Every fix, every feature, every hotfix goes through a rigorous cycle first.
For a security product where enterprises are trusting us to protect their environments, this isn't optional. It's the baseline.
!support — First Response Without the Engineer
Not every issue is a clear-cut bug. My SOC analysts often hit something ambiguous — an alert that doesn't make sense, a behaviour they can't explain, something that might be a product issue or might be a configuration problem or might be nothing at all.
Before E.D.I.T.H, that meant pinging me. I'd stop what I was doing, read the context, ask a few questions, figure out where it belonged, and route it. 10–15 minutes minimum, more if I needed to dig in.
Now they type !support <ask their doubt>. E.D.I.T.H already carries the knowledge I've built up — the platform's behaviour, the common failure patterns, who owns what, and what the right immediate action is for each type of problem. It walks through the issue with the analyst, tells them what to do right now, and tells them exactly who to contact if it needs to go further.
It isn't routing for the sake of routing. It's the first response I would give, with the experience behind it that makes that response actually useful.
The analyst gets a resolution. I stay in flow. And the issue doesn't sit unanswered because I was heads-down in something else.
Work While You Sleep 💤
This is the part that still catches me off guard sometimes.
My SOC team operates across time zones. Bugs don't wait for business hours. A critical issue can surface at midnight, and in the old world that meant either the analyst waits until morning or someone wakes me up.
Now there's a third option.
The analyst types !bug at 2am. E.D.I.T.H creates the ticket, investigates the codebase, writes the fix, reviews it, and raises the pull request. By the time I open my laptop in the morning, the work is already done. There's a PR in my DM notifications with a fix that's been thought through, coded, and self-reviewed. I read it, approve it, and it's merged before my first coffee.
I didn't lose a night. I gained one.
This changes something fundamental about how a small team can operate. You're no longer limited by the hours any one person is available. E.D.I.T.H doesn't sleep, doesn't lose context, and doesn't get pulled away by something else. It just works — steadily, reliably, in the background — while the rest of us are living our lives.
The platform keeps moving. The team keeps shipping. And I wake up to progress instead of a queue.
The Deeper Idea 💡
My SOC team now has a capable first response that doesn't route through me. My focused work blocks are longer. The low-grade anxiety of wondering what I'm missing while I'm in flow — that's gone.
My team moves differently now. They're not waiting. They have access to things that used to require a conversation with me, and that changes the dynamic in subtle but meaningful ways.
They're faster. And because they're faster, so am I.
If you're a solo or small-team engineer with people depending on you, the bottleneck isn't your capability — it's your availability. You can't be in ten places at once. But a system built on your tools, your context, and your way of working can be.
The goal was never to work more hours. It was to spend every hour on what only I can do.
E.D.I.T.H is built on Claude by Anthropic, integrated with Jira, GitHub, and Slack. It runs as a persistent bot with role-based permissions, skill-based workflows, prompt injection hardening on every user-facing skill, and a full notification layer keeping the engineer informed at every step.
~ S.G.R