AI-specifc Articles & Blogposts
https://www.ssp.sh/brain/will-ai-replace-humans/
Incredible theory and infodump by Simon Spati, bias towards negative effects but notes where it will help.
https://robertgreiner.com/believe-the-checkbook/
Follow the money and actions versus hearing & reading their words - Anthropic's acquisition of Bun (the all-in-one JS toolkit) defeats the notion that AI will be writing "all the code".
https://ploum.net/2025-12-19-prepare-for-that-world.html
Another article about looking at what is vs what is being told to you by AI companies. Prepare for the "stupid world" that will be chock-full of AI mistakes that will need to be fixed and AI gambles that will need to be recouped from.
https://tidyfirst.substack.com/p/programming-deflation
Kent Beck is middle-of-the-road on AI: it will be useful, but it will not devalue the notion of having a human write code. "Embrace the commodity and use it to do the easy stuff, focus on integrating all these AI components together, develop product/code 'taste', it's always been thinking about systems"
https://simonwillison.net/2025/Oct/7/vibe-engineering/
AI, agentic development, etc., will reward good engineering practices and systems that are made and maintained by human engineers. "Vibe" being treated like a meme by younger engineers will be their downfall, but "vibe engineering" will represent what the older engineers establish as AI initiatives gain momentum.
https://www.chrbutler.com/what-ai-is-really-for
"Doompill" take on AI - the best case is the market behind AI is a bubble and it pops as bad as the dotcom bubble. The worst case is that it normalizes and the profiteers behind it know exactly what they're doing by using it to smother everything.
https://martinalderson.com/posts/has-the-cost-of-software-just-dropped-90-percent/
Bullish view on AI: the power it gives to experienced consultants and corporate developers will reduce the cost of doing things. However, somewhat naïve view because Martin's entire opinion stems from anecdata about a handful of single large tasks he's used AI alongside.
Important note is that he believes, like Jeffrey Palermo, that AI will advance enough within the next 5-10 years that it will make singular developers much better at what they do, driving down costs for features. The HackerNews thread tends to disagree, or at least describe who exactly is getting rewarded by AI (very senior & staff engineers): https://news.ycombinator.com/item?id=46196228
https://gricha.dev/blog/the-highest-quality-codebase
Funny but insightful blog post about letting an AI agent iterate on the same codebase over and over again. AI is dangerous for a codebase's quality if overused. Claude tried to be Fabrice Bellard and roll everything itself, creating dozens and dozens of small utilities. Fabrice knows why he's rolling his own, Claude is just doing things. Test coverage skyrockets, but unless developers deep-dive into Claude's tests, we have no idea how much value those tests actually give us.
https://simonwillison.net/2025/Dec/18/code-proven-to-work/
Probably the biggest headliner: "Your job is to deliver code you have proven to work". If AI is just going to go to code review un-reviewed by the developer using the AI, or even worse AI is being allowed to push code up for review, it's going to put a massive burden on developers in terms of reviewing code. This will eradicate any benefit that AI offers developers.
https://simonwillison.net/2025/Dec/10/html-tools/
Use HTML tools as a playground for learning AI - make simple "does one thing" applications, with all CSS in the <head> tags and all JS in the <script> tag.
https://lewiscampbell.tech/blog/260114.html
Good read, highlights the continually growing backlash towards "AI Hype" and how proponents of it are becoming zealous to the point of disparagement of those who disagree.
https://www.approachwithalacrity.com/claude-ne/
"Claude is not a senior engineer - yet". Despite all the bad press, there's still some good attached to LLMs and AI for things like software development. Key sentence: "But when you don’t have good abstractions — like in our gnarly React code — Claude gets lost, and it can’t rescue itself."
https://www.caimito.net/en/blog/2025/12/07/the-recurring-dream-of-replacing-developers.html
tl;dr - We've been here before and the now-wise old tech sector folks know it. COBOL during the 70s, CASE Tools in the 80s, Visual Basic drag-and-drop UIs and Delphi in the 90s, and we've just past web frameworks and "low/no code solutions" in the 2000s. AI has a lot of potential much like COBOL, Visual Basic and Delphi, the web frameworks, etc. did, but the thinking all developers do will not be replaceable. "The artifacts we create—whether COBOL programs, Delphi forms, or Python scripts—are the visible outcome of invisible reasoning about complexity."