There’s a new rhythm emerging in digital work: running not just one coding task, but orchestrating a whole swarm of parallel agents, continuously. These are helpers, automations, tooling, and understanding AI in a way that allows to tackle problems and clear bottlenecks alongside you, based on specific intent.
These are some of my thoughts on going beyond Multitasking.
Imagine your daily workflow not as a just a to-do list, but as a team sport. One script is cleaning up warnings in your repo, another agent’s on deployment-watch in Gitea, and a language model is cranking out proof-of-concept answers or hunting for gaps in your documentation.
It turns out real productivity isn’t about competing with the raw speed of code-generating LLMs; it’s about spotting which tasks to delegate in parallel, and building ways to do it effectively:
- Research for proof-of-concepts: Fire off agents to explore libraries (even new ones missing from official docs) and to synthesize recommendations, without impacting your mainline work.
- On-demand codebase explanations: Use AI agents to map out your code: where session cookies are handled, or how legacy modules interact, etc. Save these stashed explanations to reuse for deeper context. This an absolutely change your workflows, and its the first step to go to achieve LLM generated code with quality comparable to a human engineer.
- Low-stakes maintenance: When warnings pop up or fiddly upgrades nag for attention, pass them off to bots so you can stay focused on strategic sprints.
- Directed real work: The more you specify your needs up front for an agent, up and down to the goal and approach, the less review fatigue you’ll have when merging results.
Patterns In Practice: My Own Workflows
A typical day starts with a tab group for productivity: email, calendar, code review, social, and have a sweep at the important topics. Ideally I want them all ready via one click. How do I do this now? I’m intentionally launching agents for:
- Coordinating using n8n bots to read and compile from a variety of sources, and flagging important topics. Gather all my emails, RSS feeds and unread messages, and streamlining status updates and archiving with zero context-switching. My morning brief includes the things that are deeply important to me, linked directly from a single page.
- Running multiple LLM agents (Claude, Codex) for parallel tasks—especially research or proof-of-concept investigations. I’m setting up a way to run containerized autonomous agents from my servers and personal computers, with the added capabilities of being able to access my private context automatically and act as myself on publishing and updating code, as well as controlling them while on the move (might write a post about this some day!)
- Trusting maintenance duties (like dependency upgrades) to asynchronous agents, freeing up my attention for what really matters. I can then have agents reviewing and iterating on upgrades until at least some sort of automated testing can pass. This makes me act less as a coder and more as an overseer, multiplying my output and allowing me to work at multiple projects faster by creating pipelines that keep the work going while my focus is somewhere else.
What do I need to do to “level Up”?
I’m gradually laying down my own “team playbook”, tracking what each automation or AI agent does best, and finding new steps to automate. Basically, I’ll keep evolving the setup.
Start small: Hand a bot a minor maintenance problem. See what it solves and where workflow friction still exists. Iterate, explore, and try things out, as the AI world is evolving at relentless pace, tasks that were not possible last week could not be in the realm of possibility.
Don’t conflate parallel agents with chaotic multitasking. Instead, use them to de-risk, deliver documentation, and experiment with new ideas safely and asynchronously.
Automate the grind and let your agents handle the noise, so your brain can stay on the things that matter.
Where are we going
I’m actually very excited to evolve and really putting in practice the old saying of “work smarter, not harder“. We’ve been trying this for years with some degree of success, but the latest developments feel like unblocking a new realm of possibilities. Let’s hack it away 😉
Leave a Reply