Coding tools in 2026

pacharanero · 11 February 2026 20:09

Just starting off a bit of a thread about what coding tools us Clinicians Who Code are using in 2026? I’ll start off with a few of my own.

The last year, and the last few months in particular, have seen an incredible acceleration in the capability of agentic coding tools, so my toolset has definitely shifted in that direction.

Editor

My main text editor is still VS Code. I hear a lot of people are using Cursor, and the old-skool use NeoVim of course

Terminal tools

Recently I have switched to opencode from OpenAI’s Codex CLI, because it is able to connect to a multitude of providers including GitHub Copilot, OpenAI/ChatGPT accounts, and OpenRouter via API key.

Agents

I use GitHub Copilot in VS Code, in Agent Mode. I tend to use the most recent version of GPT that I can use, at the moment that is GPT5.2-Codex, with 5.3 released in the last few days by OpenAI but not yet available in Copilot.

I also have OpenRouter accounts which enable API access to a variety of models, all through the same access credentials. Through this I tend to use Claude Opus 4.5, or (just released) 4.6.

LMStudio is a free, open source platform which allows you to run LLMs locally on your own laptop. I am experimenting with local coding agent models like Qwen 30B

Everything Digital Health

I did a few episodes of my YouTube channel Everything Digital Health covering some aspects of ‘Vibe Engineering’ or spec-driven development. I’ve put these together into a playlist for easy viewing.

https://www.youtube.com/playlist?list=PLAoA4N1oiWFAcEMAEPn-mw67fA_xKL22W

Here’s an example of a recent video

mayfield.g.kev · 12 February 2026 08:37

Do you have a default programming language, or what language do clinicians favour?

The reason I ask is I’ve started producing jupyter notebook examples using python. I believe most “developers” using our detailed data will be from analytics, life science or AI background, it seems python is the favoured language.

A EHR “developer” should be to follow these also.

ian · 12 February 2026 08:55

My current preferred language is TypeScript, but I have coded in Delphi (Pascal), Python, even a little Java and C!

I’d agree that Python is probably a great start point esp if you are interested in data + analytics but you will almost certainly find yourself needing to understand javascript for frontend work andTypeScript is a more sane version IMO!!

I use the JetBrains IDEs (WebStorm for Javascript, PyCharm for Python) as being a bit more fully featured and out-of-the-box than VSCode but I do use that for Remote Development on hosted servers as it is much better than JetBrains equivalent- no more Linux vi hell, and for ‘other stuff’. The AI support in Webstorm is very good.

I’ve toyed with Rust and Go, and I can absolutely see their value, but a high learning curve for a newbie/casual coder, and not where I would start.

pacharanero · 12 February 2026 09:20

I would say Python is a fairly safe bet. It’s being taught in high schools and is used at the highest level of machine learning, data science, and web programming. It’s solid. When I started writing “Clinicians Who Code - The Opinionated Guide” I went for Python.

My personal preference is less important - I love Ruby (although it’s so similar to Python that some valid Ruby would also be valid Python) and more recently have been getting into Rust. Discourse is Rails/Javascript so I spend a bit of time in those.

The actual language is getting less important these days though. I write specification in Markdown and I can reliably get an LLM agentic loop to build that spec in any language that is sufficiently well documented on the open web.

Hence transliterating/porting a mobile app from Swift to Dart (neither of which I know) can be done in a few days.

mayfield.g.kev · 12 February 2026 10:40

I meant to put an example of jupyter - I really do like being able to have markdown and code.

github.com/nw-gmsa/Testing

PDFTextAnalytics.ipynb

main

{
 "cells": [
  {
   "metadata": {},
   "cell_type": "markdown",
   "source": [
    "In this notebook the first section shows different libraries for reading PDF files.\n",
    "\n",
    "This was influenced by [Blueprints for Text Analytics Using Python](https://www.oreilly.com/library/view/blueprints-for-text/9781492074076/) chapters 3 `Extracting Data`\n",
    "\n",
    "The remaining sections are influenced by chapter 4 `Preparing Textual Data for Statistics and Machine Learning`\n",
    "\n",
    "## Preparing Textual Data for Statistics and Machine Learning\n",
    "\n",
    "### Extracting Data from a Laboratory Report"
   ],
   "id": "7a0852b5f09ce83a"
  },
  {
   "metadata": {

This file has been truncated. show original

I am learning NLP, python and a bit of life sciences - so the markdown (and code) isn’t that brilliant

james.roberts · 3 March 2026 12:24

Hi everyone,

My current set up is Macbook Pro M1 Max 64gb RAM.

It’s capable of running decent(ish) local models via LMstudio - GLM 4.5, qwen coder, gpt-oss-20b

I haven’t been using local models too much though as the SOTA/proprietary are generally better and faster. I tend to just use them in experimentation and for more sensitive work/documents.

I have a Claude pro subscription (I think £20/month) and use sonnet 4.6. Haiku is good for some ‘easier’ agentic work where lots of tokens are needed. I occasionally use Opus but find it maxes the limits too easily.

My setup usually uses split screen with claude code in iTerm and VScode for reviewing/editing.

Some areas I’ve been interested in recently are Neo4J and integrating these knowledge graphs with an LLM via MCP. Good courses on Graph Academy if you are interested

Adapting a pubmed MCP to improve querying of the pubmed database for research

Some useful tools I’ve found are

Pipit voice - a free local speech to text for mac. Works really fast and I set it up to use a local LLM to process the text output to auto correct and format.

I’m also looking at build better recorder - which can do local transcription and summarisation of meetings

Glad to have found this group via @pacharanero and pleased to meet you all

James

GP | RCGP Health Informatics Group, Deputy Chair | BCS PHCSG secretary

pacharanero · 3 March 2026 13:06

Nice. I have started using a similar project called Voxtype which is for Linux only (sorry Windows people - there’s gotta be something local transcription-y out there for you)

It’s built into Omarchy (an Arch Linux respin/distro/omakase customisation suite) which is how I heard about it, but I have it working on Linux Mint (Cinnamon, X11) at the moment.