completely different coding setups individuals use for programming. On this article, I’ll take you thru my private coding setup and the instruments and functions that I exploit to attain most effectivity when programming.
It is a setup that I’ve created by way of in depth testing and experimenting myself by way of trial and error. Whereas testing, I’ve tried to make use of a number of completely different functions for programming, and every of them has benefits in numerous settings.
I’ll take you thru the present coding setup I’ve, although it’s, after all, topic to vary quickly sooner or later with the fast development of LLM expertise.
I’m not sponsored by any of the tooling talked about on this video, and it’s merely the tooling I exploit on a day-to-day foundation as a programmer.
Why optimize your coding setup
As a programmer, your coding setup is without doubt one of the most necessary parts you may optimize. That is the place you spend most of your time fixing completely different issues. Due to on a regular basis you spend together with your coding setup, it’s best to spend time ensuring it’s optimized on your private workflows.
Personally, I all the time search for alternatives to make my setup extra environment friendly. For an extended time frame, I used Cursor each day because the platform from which I did all my coding. A number of weeks in the past, I abruptly shifted to utilizing purely Claude Code by way of Warp, which basically makes up the vast majority of my coding setup.
The swap from utilizing Cursor to utilizing Claude Code by way of Warp was one of the vital important productiveness will increase I’ve skilled since I first observed how environment friendly brokers might program for me. Warp + Claude Code has helped me massively in my every day work as an information scientist in a startup engaged on doc AI.
Strolling by way of my coding setup
On this part, I’ll stroll you thru the completely different tooling, strategies, and approaches I exploit to optimize my coding setup. I’ll cowl the functions I exploit on a day-to-day foundation, but in addition how I make the most of and get essentially the most out of those functions and different necessary methods I exploit to make my coding as efficient as doable.
All the suggestions I’ll cowl on this part have a major affect on my productiveness as an engineer.
Tooling
To begin with, I need to cowl the tooling I exploit. I exploit Claude Code and Warp for nearly all of my coding. If I need to test some manufacturing logs or if I need to repair a bug or implement a brand new function, I’ll basically all the time use Claude Code in Warp.
Inside Warp, I’ve the next setup. I’ve every tab in Warp as a separate folder I’m working in. So if I’m working in folder A, that’s my first tab in Warp. And if I’m working in folder B, that’s my second tab. Now, I sometimes discover myself having a number of brokers working throughout the identical folder. On this case, I make a cut up pane utilizing CMD + D in Warp, so my tab is cut up into a number of panes. Relying on the duty I’m engaged on, I might have as much as 5 brokers working throughout the identical repository. After which I’ve completely different repositories in numerous Warp tabs.
I need to word one exception the place I exploit Cursor as a substitute of Claude Code: Once I want full management of the code. For instance, if the function is of important significance or part of important infrastructure. Additionally, sometimes after I run necessary migration scripts or backfilling scripts, I’ll additionally do it in Cursor as a result of this provides me extra management of the code. I may run the code myself by way of interactive home windows with Python.
Git worktrees
As I discussed in my earlier part, I usually discover myself working a number of brokers throughout the identical repository. When you have a number of brokers updating recordsdata on the identical time in the identical repository, you’ll run into issues with brokers colliding with one another. To unravel this downside, you should utilize Git Worktrees.
Git worktrees are basically copies of Git repositories you can make to have brokers run fully separate from one another. So every time I spin up a brand new agent, I inform it to begin a brand new git worktree for what it’s engaged on. and that agent can now work fully individually from all different brokers working in the identical repository.
That is a vital function if you wish to work with parallel brokers in Claude Code (which is without doubt one of the main advantages of working with Claude Code). Thus, it’s best to positively be using Git Worktrees in your day-to-day programming with parallel brokers.
Slash instructions
Slash instructions are one other very highly effective function. Slah instructions are basically saved prompts, so you may shortly entry a immediate that you’ve saved on a earlier event. For instance, you probably have a really repetitive immediate, it’s best to retailer it as a slash command. Some examples of this are:
Slash instructions are extremely highly effective, and I’ve coated them in considered one of my earlier articles. The advantage of slash instructions is twofold. To begin with, you save time by not having to write down out the immediate each time. So as a substitute of getting to write down out an extended immediate, telling the mannequin that it must:
- Pull the most recent dev department and rebase on prime of it
- Run precommit checks
- A superb PR description
- Make a pull request from a function department to dev
As an alternative of getting to write down out all of this, you may merely retailer this immediate in a slash command and entry the immediate immediately.
The second benefit is that you simply get to be extra constant when writing your prompts. For instance, when creating pull requests to dev, as I discussed, you would need to run a collection of checks (pull newest dev department, rebase, run precommit checks, …). In the event you write this out each time, you danger forgetting components of the immediate. This isn’t an issue if you happen to use slash instructions, nevertheless, since you’ll all the time be using the identical immediate, and also you’ll be extra constant.
Slash instructions make you each quicker and extra constant
Low threshold to fireside off brokers
One other matter I need to cowl is that it’s best to have a brilliant low threshold to fireside off brokers to carry out duties for you. At any time when you may consider a brand new activity or get a brand new downside it’s important to clear up, it’s best to simply hearth off an agent. For instance, if I discover a button that’s misaligned, some textual content in my software that needs to be up to date, or translations that should be up to date. I merely hearth off a brand new agent, let it run totally autonomously, and create a pull request for me.
The principle level is that it’s best to have a low threshold to fireside off brokers as a result of it’s so low-cost to run and prices you so little time. The price of firing a brand new agent is basically spending time writing out immediate and, in lots of instances, answering just a few questions the brokers have so that you can correctly perceive the duty you gave them.
There at the moment are many instruments on the market that supply a number of token utilization for a comparatively low value. For instance, I’m utilizing the $200 Claude Code subscription, which is a set quantity per 30 days, and I’ve by no means run into price limits. This implies I can hearth off as many brokers as I can with out extra price.
Make the most of the very best fashions
One other tip I’ve may sound very apparent, however I all the time advocate utilizing the very best fashions everytime you work with programming. The rationale for that is that in the long run, this protects you each money and time.
Sure, the very best fashions are sometimes the most costly fashions per token and are additionally the slowest fashions. Nevertheless, it seems that if you happen to use cheaper fashions, they are going to extra usually make errors, which takes extra time so that you can repair and iterate on, which once more makes the mannequin make the most of much more tokens. Thus, in the long run, it usually seems that utilizing a less expensive, smaller mannequin truly seems to be dearer and time-consuming.
It is best to due to this fact be using the frontier fashions resembling Gemini 3 Professional, Claude 4.5 Opus, and GPT 5.2 Codex. There are additionally some up-and-coming open supply fashions performing properly on the coding benchmarks, although I haven’t achieved the identical success with open supply fashions as I’ve achieved with frontier closed supply fashions.
Conclusion
On this article, I’ve coated how you can have a most effectivity coding setup. I’ve mentioned the coding setup I exploit on a day-to-day foundation, the place I exploit the Warp terminal with Claude Code. Moreover, I exploit particular strategies resembling organizing Warp with cut up panes and tabs by the folder I’m engaged on. I’m additionally ensuring to all the time use the most recent and finest coding fashions. I consider spending time optimizing your coding setup is an excellent use of time. As a programmer, your coding setup is without doubt one of the belongings you spend essentially the most time with, and if you can also make that just a few proportion factors extra environment friendly, it should possible repay in the long term.
👉 My Free Sources
💌 Substack
🚀 10x Your Engineering with LLMs (Free 3-Day Email Course)
📚 Get my free Vision Language Models ebook
💻 My webinar on Vision Language Models
👉 Discover me on socials:
🧑💻 Get in touch
✍️ Medium
