Home
AI
Blog

Issue tracking is dead? The shift to context + agents

basanta sapkota

“Issue tracking is dead” is the kind of headline that makes an engineer pause mid-command and squint at the screen. Really? Because most teams I know still basically live in Jira, Linear, or GitHub Issues.

But Linear’s point isn’t “bugs are gone, tasks are gone, party’s over.” It’s this: the job issue tracking used to do is getting eaten alive by AI agents, especially the parts are pure ceremony. The handoffs.Now status ping-pong.Now endless ticket grooming where everyone nods and nobody remembers what they just agreed to.

And yeah… spicy take. It’s supposed to be.

The quick takeaways, in plain English

  • When people say “issue tracking is dead,” what they usually mean is handoff-heavy workflows are dying. Accountability isn’t dying. Bug tracking isn’t dying.
  • Linear says coding agents are installed in 75%+ of their enterprise workspaces. They also say agent-completed work grew 5× in three months, and agents authored around 25% of new issues. That’s not a rounding error.Now’s a shift in who, or what, is producing the tickets.
  • The new choke point is context. Decisions, constraints, customer pain, why we’re doing it this way, what broke last time… the stuff that used to be in someone’s head or scattered across docs and chat threads.
  • Strong teams will keep issues, but treat them like an execution artifact, not the place where product thinking magically happens.
  • If you want a healthier scoreboard, look at DORA metrics: lead time, deployment frequency, change fail rate, recovery time. Ticket throughput is easy to game.

What “issue tracking is dead” actually means… and what it doesn’t

Here’s the clean version.

Not dead:
Tracking bugs. Prioritizing work. Knowing what shipped. Auditing decisions. Having a system of record.

On the way out:
Using ticket rituals as a substitute for shared understanding.

Linear’s argument is basically that issue tracking was built for a handoff model. PM scopes now, engineering picks up later, the workflow tries to bridge the gap… and eventually the process becomes the work.

Their Next page doesn’t mince words. Agents compress planning, implementation, and review, so the system should be designed around context and agents, not a parade of statuses. They position Linear as “the shared product system that turns context into execution.”
External link: https://linear.app/next

If you’ve ever spent a whole sprint grooming tickets nobody reads, you already know what nerve they’re touching.

No. Issue tracking isn’t dead.

But “issue tracking is dead” works as a provocation because teams are drifting away from status-driven ticket workflows and toward context-first systems, where agents handle more of the clerical work.

Why this is blowing up right now: agents + context

Linear actually brought receipts. Their adoption numbers:

  • Agents installed in more than 75% of Linear’s enterprise workspaces
  • Over the last three months, work completed by agents grew 5×
  • Agents authored nearly 25% of new issues

Source: https://linear.app/next

Those numbers matter because they hint at a new default: not “humans write tickets for humans,” but “humans and systems provide context so agents can act.”

A Reddit thread reacting to Linear’s post puts it in a way feels painfully true. The bottleneck is shifting from execution to cross-agent context. A lot of the real decisions happen in ChatGPT or Claude chats, random docs, and messy debates before anything becomes an issue.
Source: https://www.reddit.com/r/Linear/comments/1s2mf7t/linears_issue_tracking_is_dead_post_makes_me/

Then Angela Hottinger on LinkedIn echoed the same direction. The next era is context and orchestration, not handoffs and status updates.
Source: https://www.linkedin.com/posts/angelahottinger_issue-tracking-is-dead-the-next-era-of-activity-7442290736462573568-Vsow

Issue tracking isn’t dead for bugs, and Jira’s definition still fits

Let’s not pretend bug tracking is going anywhere. It’s not.

Atlassian’s definition is still the day-to-day reality. A centralized system to find, track, report, assign, and keep a “single source of truth” for backlog items.
External link: https://www.atlassian.com/software/jira/features/bug-tracking

What changes is what “single source of truth” needs to contain. It can’t just be status, points, assignee.Yet has to hold the real fuel agents and humans run on:

  • customer feedback and user impact
  • constraints and decisions, the why behind the work
  • links to code, logs, incidents, rollbacks
  • what the agent tried, what it changed, and what it couldn’t decide

That’s not a classic ticket anymore. It’s closer to a little case file.

If issue tracking fades, what replaces it? A context-first workflow

Linear literally shows the pipeline as “Context → Rules/Automations/Skills → Agents.” If you’re adding an image to your post, that’s the one.

How it tends to look in real life, with all the bumps and duct tape:

  1. Capture context near where it happens. Support inbox. Slack thread. Incident channel.
  2. Normalize it. Summaries, decisions, acceptance criteria.
  3. Route it. Automation and triage to the right owner or agent.
  4. Execute. Agent drafts code or a PR, human reviews for judgment and taste.
  5. Measure outcomes. Did failures go down?But lead time improve? Not “did we close 37 tickets.”

Keeping issue tracking useful in an agentic world: stuff actually helps

Tickets start feeling like busywork when they stop carrying meaning. I’ve been there. You open an issue and it’s basically “Fix login. Priority: High.” Cool. High for who, and based on what?

Here are a few moves keep the system from turning into a graveyard.

1) Turn issues into “context packets,” not status containers

Minimum fields I like to see:

  • problem statement, what breaks and who it hurts
  • evidence, logs, screenshots, traces
  • decision record, why this approach and what constraints shaped it
  • links, PR, docs, incident

On GitHub, issue forms can help enforce this structure:

# .github/ISSUE_TEMPLATE/bug.yml
name. Bug report
description. Track a bug with context agents can use
body.
  - type. Textarea
    id. Problem
    attributes. Label. Problem
      description. What’s happening? Who is impacted?
    validations. Required. True
  - type. Textarea
    id. Evidence
    attributes. Label. Evidence
      description. Logs, screenshots, traces, links
  - type. Textarea
    id. Decision
    attributes. Label: Decision/constraints
      description: Relevant tradeoffs, constraints, and why

2) Stop using “tickets closed” as a productivity religion

DORA’s delivery metrics are a better north star because they reflect speed and stability. change lead time, deployment frequency, change fail rate, recovery time. Their current guide also includes deployment rework rate.
External link: https://dora.dev/guides/dora-metrics/

And yes, it matches the Agile Manifesto vibe. “Individuals and interactions over processes and tools.” Tools matter. They’re just not the point.
Source: https://agilemanifesto.org/

3) Automate triage, not judgment

The Register’s write-up on Linear’s agent launch calls out “skills” for saved workflows and “automations” for triggered workflows. It also points at where things get sharp fast. Prompt injection and permission boundaries become a bigger deal as agents gain power.
Source: https://www.theregister.com/2026/03/26/linear_agent/

My rule of thumb: let automation classify and route, but keep humans on the hook for priority, user-impact tradeoffs, and the final “ship it” call.

Common ways teams mess this up when they declare “issue tracking is dead”

I’ve watched orgs try to “move past tickets” and accidentally reinvent Jira inside Slack. Same chaos, just louder.

A few traps worth dodging:

  • No written decisions, so context evaporates and nobody remembers why two months later.
  • Agents with no guardrails. If an agent can open PRs, define what “done” means and who reviews.
  • Replacing issues with vibes. You still need a system of record for bugs, incidents, commitments.
  • Tracking everything except outcomes. If lead time and change fail rate get worse, the shiny workflow isn’t working.

So… is issue tracking dead?

Half true. The best hot takes usually are.So “dead” part is the handoff-era obsession with process-as-progress.Still “alive” part is the need to track problems, decisions, and delivery in a way people can trust.

If you’re playing with coding agents right now, try one small experiment this week: rewrite one recurring ticket type into a context packet with problem, evidence, decision, links. See if it cuts the back-and-forth.

And if you want more agentic dev tooling ideas, I recently rounded up a bunch here: https://www.basantasapkota026.com.np/2026/03/top-10-agentic-coding-tools-in-2026-dev.html
Drop a comment with what’s actually breaking in your workflow. Tickets, context, reviews, any of it. I’m genuinely curious.

Sources

Post a Comment