AI promised less work. I've never worked more
TL;DR Highlight
Running agents in parallel brought my GitHub activity back to full-time engineer levels, and I'm exhausted.
Who Should Read
A developer or engineering lead running multiple coding agents in parallel who feels context-switching fatigue and burnout
Core Mechanics
- Work didn't shrink, it multiplied
- He used to do two or three serious tasks a week, one at a time.
- Now several agents run implementation, investigation, review, and research in parallel.
- The problem is the agents' 'sometimes'
- Each agent eventually returns with a question, a plan to approve, a diff, or a result.
- Every reply means reloading a different repo and a different set of decisions.
- GitHub history is back at Full Stack levels
- Median weekday was 3-4 contributions in 2022-2025 and is 7 in 2026 (8 in 2020).
- Days with 24+ contributions were 1 per year in 2022-2025 and are already 16 in 2026 (17 in all of 2020).
- More activity despite less coding in the role
- By Oct 27, 2026 he has 2,102 contributions, more than any full year from 2021 to 2025.
- Same period in 2020 had 2,371.
- Cheap implementation grew the wish list
- He now fixes bugs he'd have left alone, improves patterns, and starts new projects.
- Agents didn't just speed up the old workload, they made a much bigger one easy to take on.
- Bursts then burnout, repeatedly
- A new model or workflow floods him with ideas, so he starts more agents.
- Reviewing and context switching catch up, and he counts three cycles since January.
- August hit 427 contributions, versus the previous high of 319 in April 2020.
Community Reactions
- A daily fixed to-do list worked
- A commenter clocked out earlier and slept better on days they kept the list.
- Days they lost control of it brought exhaustion the same day; they suggest planning changes on paper without the LLM.
- Scaling up agents to reduce load
- Two commenters (father and sons) each manage about 100 agents and talk to only about five coordinators.
- They say this gives room to think, pause, and focus again.
- Drudge work was brain downtime, one opinion says
- Writing code to spec is cognitively light and gave rest between intense thinking phases.
- With AI doing it, only the hard part remains, hence more stress.
- A commenter swapped their smartphone for a flip phone
- Having agents on the phone meant every thought became a prompt, so work never stopped.
- Now work happens mostly at the computer during work hours.
- Many say employers just demand more
- Several commenters argue tools never reduce work, only raise expectations.
- Jevons paradox was cited for management expectations rising.
How to Apply
- Fix today's task list before starting
- Right after you start work, write at most 5 tasks for today on paper or a note.
- Write off-list ideas in a 'later' column without launching an agent.
- Cap concurrent agents
- This week, limit simultaneous agents to 2-3 and log context switches at day's end.
- Lower the cap if you're drained, raise by one if you're fine.
- Batch your agent check-ins
- Handle questions, approvals, and diffs 2-3 times a day at set times.
- Mute notifications and don't open agent replies outside those slots.
- Gate extra projects behind a planning session
- Prioritize new ideas on paper without the LLM before starting them.
- Remove one existing item when you add a new one to keep the total constant.
Try It
# Paste into Claude or ChatGPT (a simplified version of the article's observation: parallel agent load check)
I'm a developer running several coding agents in parallel. Here are today's tasks and agents.
Today's tasks: {{today's task list}}
Current agents/tasks: {{tasks per agent}}
Focus time available today: {{hours}}
Please do the following.
1. Point out which task combinations will cause the most context switching.
2. Pick at most 3 tasks that I can safely run at once today.
3. Move the rest to a 'later' list with a one-line reason each.
4. Schedule agent check-ins in 3 slots across the day.
Next step: at day's end, send me what leaked outside today's list and ask for a cap adjustment for tomorrow.Terminology
Related Papers
The Agent SDK
Cursor released Agent SDK, letting you define agents in code and deploy them with one command.
Here be monsters: What do you still need to know when agents do the work?
Owning agent-written code depends on fundamentals, breadth, and diving deep just in time.
Show HN: Corgi, a helpful {agent, terminal} multiplexer
A tmux-like terminal tool that shows the status of all your AI agents at a glance
Ask HN: If the models are so good, why doesn't OpenAI just start 100 companies?
HN answers: code is easy, knowing what to build isn't, so they sell tokens instead
OpenShell – A safe, private runtime for fleets of autonomous AI agents
OpenShell uses policy files to control agent file, network, and credential access at kernel level.
Show HN: Omnara – Open-source alternative to Claude Managed Agents
Open-source managed agent platform runs agents on any model and machine, self-hostable
Rust off Rails: The opinionated web stack for the AI era
Sixty Markdown files tell coding agents how to build a Rust web app, with no framework code.
LayerFS: The filesystem for AI coding agents
Mount an ephemeral workspace per agent, commit only the changes