In a previous article, I of an important techniques I utilize to code effectively with AI agents. In this text, I’m continuing with 4 additional techniques, all of which I exploit each day.
I imagine that with a purpose to be an efficient programmer today, you could have to heavily utilize AI tools. When you’re not coding using AI agents, you’re falling behind. Moreover, agents will be used for therefore rather more than coding as well:
- Agents can read and create Linear issues
- Agents can perform deep research on a subject you’re serious about
- Agents can review log messages from production code
All of that are vital tasks, programmers need to perform regularly.
Thus, I’m advocating for heavy use of AI agents to be as efficient as possible. In this text, I’ll cover, on a high level, 4 more techniques I utilize that I imagine make me a more efficient programmer.
- Macwhisper for faster agent prompting
- Claude Code review
- Parallel agents
- Interacting with GitHub using agents
I’m also very serious about hearing if you could have any techniques which are vital in your programming workflows. If you could have specific techniques in mind, be at liberty to achieve out, as I’d love to listen to about it.
Why it’s best to code with AI agents
I’ve previously described how coding with AI agents make me so much more practical as a programmer. I’ve multiplied my programming output over and over through using AI, and it simply allows me to do rather more than I did previously.
A typical counterargument to AI agents is that you want to understand your code before pushing it to production. I agree with this assessment to some extent in the event you’re working with critical systems which are hard to perform end-to-end tests on.
Nevertheless, most web sites and applications usually are not like this. To start with, they’re not as critical, and secondly, most tasks you’re employed on as a programmer are verifiable. This implies you possibly can often test behaviour just by literally testing if the feature works whenever you interact with it.
Thus, I’m advocating for more use of AI agents and for using them for all programming-related tasks. For instance:
- Create Linear issues
- Fix bugs by simply linking to the Linear issue
- Planning and developing latest features
4 Techniques for coding efficiency
On this section, I’ll cover 4 techniques that I exploit for my AI-native programming workflows. These are specific techniques that I literally use daily I program.
Macwhisper
MacWhisper is an excellent transcription tool available on Mac. Simply put, Macwhisper lets you press a button, confer with your computer, and the text is routinely transcribed and pasted wherever your mouse cursor is.
This is useful because numerous my programming workflows have moved from pure code to natural language. Using a transcription tool for coding would naturally be hard because coding requires numerous special characters like colons, parentheses, and tabs, that are faster to type on a keyboard.
With AI agents, increasingly work is finished in natural langauge, as an alternative of coding language.
Thus, each time I prompt my Cursor agent, I often just hold down the button and say out loud whatever I need to prompt my agent. I’d, for instance, ask:
Check the logs for this document id, was it processed accurately In this instance, I paste within the document id after saying the sentence out loud.
The rationale I exploit Macwhisper is solely that I talk faster than I can type. The typical talking speed is around 150 words per minute, while most individuals can’t type 100 words per minute at maximum speed. Moreover, you’re rarely capable of type at max speed when you could have to think as well.
Claude Code review
This step is split into two parts:
- After implementing a feature, I ask Cursor if the code is production-ready, and only push when Cursor is satisfied
- Each time I make a PR, I even have Claude Code to a code review as well, with no other context than the PR description, and the Git diff file to the branch I’m merging to.
This works thoroughly. Asking Cursor if the code is production-ready makes Cursor do a review of my changes and fix any small issues that may not work as intended.
Moreover, having a totally separate LLM review the code with no context of how the implementation was done is super helpful. This often discovers other errors that I (or Cursor) didn’t take into consideration when implementing the code within the PR. This also significantly lowers the quantity of bugs experienced in production, and is a comparatively low-cost addition you possibly can make to your CICD pipeline.
Parallel agents (fire and forget)
One other vital technique is to make use of parallel agents. Each time I’m blocked by an agent doing a little work, I at all times start a brand new agent. This may very well be a coding agent implementing one other feature, or it may very well be Gemini deep research, researching a subject I’m serious about. The purpose is that I never simply wait on my agent without doing anything.
When running parallel agents, you would possibly start fighting context switching. Switching contexts often could be very taxing on your brain, and is unquestionably something you would like to minimize.
Thus, I at all times be sure that I work on a task until I’m fully blocked. I try to attenuate the variety of times I switch context, and only start a parallel task once I’m sure I even have to attend a while for my coding agent to complete its implementation.
One other vital point here is that you simply give your coding agents enough permissions to run for an prolonged time frame. When you’re interrupted on a regular basis with the cursor asking for permissions, the parallel workflow doesn’t work well.
You may have to provide your coding agents enough permissions. When you’re at all times interrupted with a permission request, it’s hard to work effectively.
Commit and PR with agents
Lastly, I need to focus on how I at all times interact with GitHub using my coding agents, as an alternative of writing the commands myself. The rationale I do that is that it’s simply faster, and I can do something else while my agent runs precommit hooks, commits, pushes, and makes pull requests.
Writing commit messages, pull request titles, and descriptions takes a surprising period of time. Especially whenever you’re performing quick actions, comparable to adding translations or moving a button within the UI. Subsequently, I at all times utilize Claude to jot down my commit messages, PR titles, and descriptions.
Not only does this save me time, but I also think Claude does a greater job at writing these messages for me. With pull requests, for instance, it’s often hard for a human to recollect all the changes made and to summarize them in a pleasant manner. It’s much easier for Claude to take a look at the Git diff and supply a summary of all of the changes made.
Thus, I’ve given Cursor permission to interact with GitHub for me. As an alternative of performing all the GitHub actions myself, comparable to:
- Pulling
- Rebasing
- Commiting
- Amending
- Pushing
- Creating PR’s
I simply prompt Cursor to do it for me. Thus, I can just fire and forget. My workflow after implementing a brand new feature is solely to offer the next prompt to Cursor:
Run all precommit checks (black, mypy, pytest), commit and push.
Then create a PR on this branch and provides me
the link to the PR That is a lot faster than writing the GitHub commands yourself. Not having to jot down pull requests myself might be an important bit, as this was super time-consuming previously after I made my pull requests within the GitHub UI. Now I simply click the link my agent provides me, and the PR is prepared. Then I take a look at the Claude Code review provided to me, and fix any potential issues.
Conclusion
In this text, I’ve discussed 4 specific techniques I exploit each day after I’m coding. I discussed Macwhisper for transcription, Claude Code reviews, parallel agents, and interacting with GitHub using my agent. Together, I estimate that these techniques save me at the least 1 hour every day, which is a big period of time. Freeing up this time allows me to finish so many more tasks over the course of a project. I imagine that being effective with AI agents is an excellent vital skill, and definitely a subject it’s best to spend time becoming good at.
👉 My Free Resources
🚀 10x Your Engineering with LLMs (Free 3-Day Email Course)
📚 Get my free Vision Language Models ebook
💻 My webinar on Vision Language Models
👉 Find me on socials:
🧑💻 Get in contact
✍️ Medium
