On March 24th, 2026, Linear published a page that changed how I think about the next decade of software. Under the heading "Issue tracking is dead," they announced that Linear is no longer primarily a tool for humans. It's becoming a shared system for humans and agents.
Their reasoning is sound. In 75% of Linear's enterprise workspaces, agents are already installed. 25% of new issues are created by agents. The trajectory is clear: within two years, most issues will be created, triaged, and executed by agents. The human's role shifts from doing the work to defining and reviewing it.
Linear is adapting to this reality. They're right to do so. But their announcement clarified something else — something about the work that remains after agents take over execution.
#The unit of work is splitting
Linear's atomic unit is the issue — a discrete, movable work item. Write a function. Fix a bug. Update a dependency. An issue has a clear definition, a clear completion state, and increasingly, a clear path to automation.
Particle's atomic unit is the session — the span of time in which a human is present, cognitively engaged, and spending energy on decisions that matter. A session can't be automated. It can't be delegated. It can't be parallelized across eight agent instances.
This distinction didn't matter much when humans did both: the thinking and the executing. You'd spend four hours coding a feature — two hours understanding the problem, two hours writing the solution. The thinking and the doing were interleaved, inseparable.
That's ending. The doing is being separated from the thinking. And the doing is going to agents.
#What remains
Here's what an agent can't do:
It can't decide that this feature shouldn't be built at all. It can't feel that a design is off by one degree — not wrong, just not right. It can't sense that a customer's frustration isn't about the bug they reported but about the workflow that led to it. It can't sit in a room and read the unspoken tension about a deadline that everyone knows is unrealistic.
An agent can execute a decision in seconds. But it can't make the decision. Not the ones that matter.
The decisions that matter are the ones where reasonable people would disagree. Ship this feature now or wait for better data? Simplify the interface at the cost of power users? Hire for speed or for culture? Apologize publicly or handle it privately?
These are judgment calls. They require context that doesn't fit in a prompt. They require taste that can't be fine-tuned. They require the accumulated weight of experience that comes from having sat in rooms, shipped products, made mistakes, and learned from them.
This is what humans do. This is what humans will keep doing.
#The compression
But here's the part nobody is talking about: when agents handle execution, the remaining human work doesn't get easier. It gets denser.
Consider a software engineer's day in 2024. Four hours to build a feature. Maybe five decisions along the way: architecture, naming, edge cases, testing strategy, when to ship.
Now consider the same engineer in 2027. The agent builds the feature in twenty minutes. The engineer reviews it in ten. Then moves to the next one. And the next. In those same four hours, they're not making five decisions. They're making forty.
Each decision is faster. But there are more of them. And each one carries more weight, because the agent executes immediately. A bad decision at 9 AM becomes a deployed mistake by 9:15.
The cognitive load doesn't decrease. It compresses. More judgment per hour. More context-switching between domains. More moments where the answer isn't obvious and you need to think clearly.
This is the new shape of work. Not less human involvement — more concentrated human involvement. Not fewer hours — more demanding hours.
#The missing layer
Here's what's strange: we have tools for everything around this work, but nothing for the work itself.
We have tools that manage what needs to be done (Linear, Jira, Asana). We have tools that agents use to do it (Copilot, Cursor, Devin). We have tools that track what was done (Git, deployment logs, analytics dashboards).
But for the human sitting at the desk — the one making forty decisions in four hours, the one whose judgment is now the most valuable part of the entire chain — there's nothing.
No tool that helps them structure their session. No tool that tracks their cognitive rhythm. No tool that says: you've been in deep review for ninety minutes, and your decision quality drops after sixty. No tool that helps them reflect on whether they spent their morning on the right problems.
The human layer of knowledge work is entirely uninstrumented.
#Why this matters now
When execution was slow, this gap was invisible. If building a feature takes four hours, the pace is naturally regulated. You have built-in breaks. You have time to think between actions. The rhythm of execution enforces its own structure.
When execution becomes instant, that natural regulation vanishes. The agent is always ready for the next instruction. The pull request is always ready for review. The next decision is always waiting. Without deliberate structure, the human becomes the bottleneck — not because they're slow, but because they're unprotected.
This is already happening. Developers who use AI coding assistants report that they write more code but feel more exhausted. Managers whose teams ship faster find themselves in more review meetings with less time to think between them. The acceleration of execution is creating a compression of judgment that nobody designed for.
#The session as foundation
A session — a deliberate block of time with a beginning, an end, and an intention — is the oldest structure in human work. Long before software, before offices, before industrial manufacturing: people worked, and then they rested.
This rhythm isn't a productivity hack. It's physiology. Cognitive performance follows a curve. Attention is a resource that depletes and replenishes. The quality of your work in hour five depends on what happened in hour three.
What changes in the agent era isn't the session itself. It's what fills it. Instead of four hours of mixed execution and thinking, it's two hours of pure thinking — pure judgment, pure decision-making — followed by rest that actually needs to replenish something.
The session becomes more important, not less. And the quality of that session — how you enter it, what you focus on, how you end it, what you do afterward — becomes the primary determinant of your output.
Not your typing speed. Not your tool proficiency. Not your ability to hold a codebase in your head. Your ability to make good decisions, repeatedly, under the compression of agent-accelerated execution.
#What we're building
Particle started as a focus timer. A beautiful, minimal, keyboard-first way to structure sessions of deep work. Particles as a unit of focused energy. A companion for people who take their work seriously.
That foundation turns out to be exactly right for what's coming.
The Particle Loop — capture, plan, execute, complete, reflect, align — is a cycle of continuous improvement. Not for productivity. For the quality of your judgment. For learning from your own patterns. For becoming better at the work that only you can do.
When we built reflection into Particle, we weren't thinking about agents. We were thinking about craft. About the difference between someone who works and someone who gets better at working. About the closed loop that turns experience into wisdom.
It turns out that loop is precisely what the agent era demands. Agents can optimize themselves. They get better with every iteration, every fine-tuning run, every expanded training set. Humans can do the same — if they have the loop for it. If they have a space that treats their sessions not as time blocks to fill but as opportunities to grow.
#The last interface
Most of the software you use today will become an agent interface. Email will be filtered and drafted by agents. Project management will be organized by agents. Code will be written by agents.
But there will be one interface that remains entirely, irreducibly human. The interface for the moment when you sit down, set an intention, focus, decide, and emerge having done work that mattered.
That's not an interface that can be automated. It's an interface for the part of work that makes you human.
We're building that interface. Not because it's a market opportunity, though it is. Because it's the only interface that will still matter when the agents take over everything else.
The work of a lifetime consists of many particles. The agents will handle the execution. But the particles — the focused sessions where you bring your full judgment, your accumulated context, your irreplaceable taste — those are yours.
They always will be.