AI is here to stay. It has revealed how much of our daily work is repetitive, much like how we once moved away from typists and data entry clerks. The future of work is more abstract, like coding in high-level languages instead of assembly. Forced memorization and routine tasks will lose their value. What rises in value is agency, imagination, curiosity, and the ability to ask good questions. These are things AI cannot replace… yet!
I’m writing this for future reference. In the future, we can look back and see how “primitive” AI was in 2026.
Writing
Software folks, whether they work remotely or not, write plenty to communicate with fellow team members and their future selves. Spell checkers and grammar tools have existed long before AI. What’s different now is that AI also helps me shape a piece more quickly, suggesting structure and phrasing beyond surface-level corrections.
For technical writing, AI can provide a basic structure. With that in place, I fill in the blanks with information I’ve gathered. The main drawback is that AI can make writing sound generic, but for technical writing that’s often fine. For creative writing (like this blog post!), human intervention matters more. I ensure the tone stays consistent and personal.
I do not have a go-to AI for writing at the moment. This post has been aided by Notion AI, Claude Code and ChatGPT.
Gathering and Processing Information
AI has been a useful companion for research. For my upcoming trip, I gave Notion AI a list of work-related conferences I had attended and ones I was interested in attending, along with upcoming non-work events in nearby countries. It suggested other events I might enjoy and laid everything out in a table. My travel plan was 90% done.
AI also helps me learn from long-form content. For public presentations, I use ChatGPT’s microphone feature to follow along and ask questions afterwards, catching points I might have missed. For longer YouTube videos, I ask AI for a summary first to decide if it is worth watching in full, then use it to find timestamps or dig deeper into specific topics.
I like Notion AI’s chat interface working directly with the document and the microphone feature from ChatGPT web to capture information quickly. Google Docs, even with Gemini integrated, feels behind in AI integration compared to the others.
Coding
In coding, AI often feels like a glorified Stack Overflow. It’s useful for quickly researching a framework, getting syntax and library suggestions, and clarifying what a framework can and cannot do. In addition, it applies the code changes without the classic “copy and paste from Stack Overflow”.
AI has helped me with code refactoring. With the right prompts and good context, it can help me restructure code much faster than I would have done alone. (For example, PR #231 was conceived and done within an hour during a commute. đźš‚)
It also helps with troubleshooting: resolving linting issues, analyzing test results and interpreting error messages. These tasks used to involve a lot of searching. AI consolidates that into a faster feedback loop.
Claude Code and GitHub Copilot, both integrated with VS Code, have been my daily drivers for coding-related tasks, gathering context from my terminal and the codebase to provide informed answers.
What work means moving forward
Abstract and critical thinking is expected rather than optional: defining problems, making trade-offs, and composing systems, rather than cranking out code. Knowing a fact, such as reciting the definition of call-by-value, no longer suffices. Understanding how a fact affects how things work together becomes expected.
AI-related tooling should include more ways to specify context. I’ve mentioned how the existing codebase and the microphone can help shape the question in addition to the prompt. In the immediate future, using MCP sounds like a good idea to incorporate more context into the prompt.
Let me look back to this post in 1 year, 2 years, 5 years, 10 years…
