In years but all the agents work best with good static analysis tools. Might be best to convert it to typescript first...
In years but all the agents work best with good static analysis tools. Might be best to convert it to typescript first...
Probably can't do it without some kind of paid plan, but Coddex with a chatgpt subscription or Claude code with Claude subscription or copilot cli with a copilot subscription can handle this. Works best if there are good tests and some kind of lint (eslint or biome). Haven't worked with flow in
Google seems willing to spend billions on map features like Street View, but seems to be unwilling to invest a few developer days to improve its pathing algorithm so it realizes that turning left on a major street without a light might add 5-10 minutes to your drive.
I used to think “the end of coding” would arrive as a single catastrophic failure. Lately it looks quieter: agentic frameworks like OpenClaw pushing more of the craft into loops you don’t fully watch—and sometimes, code you don’t even read.
agingcoder.com/posts/agents...
#AgenticAI #LLMs #Agents
One real use case I thought of for the Copilot SDK (github.com/github/copil...) would be as a first response SRE, but it looks like not yet for even the Frontier models quesma.com/blog/introdu....
I Built an Interactive CV: a resume you can talk to, plus a Fit Assessment that compares job posts to my background. Writeup covers what it took to build it over a weekend:
Writeup: agingcoder.com/posts/i-buil...
Interactive CV: agingcoder.com/cv/
#AI #LLM #RAG
Want to see a cost breakdown and some solid modelling for how public grocery stores could work in Canada? Read a detailed new report from the Canadian Centre for Policy Alternatives:
A compass encircled with various travel imagery overtop of the image of a winding road.
People keep talking about how amazing Google's NanoBanana is about visualizing and understanding text and video, and yet when I asked it come up with an feature image for my last blogpost (agingcoder.com/posts/ai-ass...) this was the weirdness it came up with.
New post: AI-assisted coding on a budget.
Agentic coding workflows are here, but code review, tests, and budgets still matter. I wrote about building new projects under token constraints and why inference cost is a real factor.
agingcoder.com/posts/ai-ass...
#AIEngineering #Programming #LLMs
AI coding is real.
The perils are real.
The hype is… loud.
My take on Vibe Coding: useful framing, obvious lessons, lots of FAAFO repetition, unresolved contradictions between rigor and YOLO agent chaos.
agingcoder.com/posts/book-r...
#AICoding #VibeCoding #SoftwareEngineering #LLMs #DevOps
Ten years ago I called React Native a “tempting quagmire.” After building a new app in 2025, my verdict is very different. What’s changed, what still hurts, and when I’d recommend it now: agingcoder.com/posts/lookin... #ReactNative #TypeScript
To “my students and to anyone who might listen, I say: Don’t surrender to AI your ability to read, write and think when others once risked their lives and died for the freedom to do so.”
www.huffpost.com/entry/histor...
There is definitely something going on with context in ChatGPT 5.1 today. It will answer all the previous questions in the thread (with similar answers to previous results) before answering the current question leading to pages of text. Switching back to 5 Instant or 5 Thinking fixes everything. 🤔
"In August 2025, Google announced that starting next year, it will no longer be possible to develop apps for the Android platform without first registering centrally with Google." What tis means for your rights, and how you can help: keepandroidopen.org
New blog post: RAG Time — Cooking Up Smart Recipe Suggestions with AI
Used a recipe database + embeddings + cosine similarity to build a FastAPI service that suggests recipes from your leftover ingredients.
A tasty intro to RAG.
Read here: agingcoder.com/posts/rag-ti...
I made a simple web app where AIs play a telephone game by taking turns creating images and trying to guess the original text. Unfortunately they are far too good at it so you don't get the comical errors humans make. ai-telephone.timfduffy.com
If you're on #LinkedIn, this is the link for declining their use of your data for training their AI models:
www.linkedin.com/mypreference...
Follow-up to my post on clustering recipes with unsupervised learning: now I’ve built a FastAPI endpoint to actually serve recommendations!
From ML exploration → usable service.https://agingcoder.com/posts/2025-08-30-serving-the-cookbook-creating-an-endpoint-for-recipe-recommendations/
Continuing my ML series—this time unsupervised. What can we do with our recipe database? Vectorize ingredients, cluster dishes, and surface look-alikes.
🔗 agingcoder.com/posts/2025-0... #MachineLearning #UnsupervisedLearning #DataScience
Who really represents you?
Under first-past-the-post, most people are stuck with a single MP they didn’t vote for.
Proportional representation gives you a team of local or regional MPs who reflect your values.
Here’s how proportional representation strengthens local democracy 🧵1/8
Not to say everything is wonderful in ChatGPT world. I tried to have ChatGPT improve my resume and asking it to put my uploaded markdown resume into the Canvas it invented 3 jobs, added 3 papers I had never written (there are no papers in my CV) and added a talk I had never given (no talks either)
Kind of sad, Gemini doesn't know anything about Google's own applications. But ChatGPT had no problem answering.
ikea captcha
Duolingo share price since their "AI first" disaster - down ~35%
Daily active user (DAU) growth down 25%
Can a tiny LLM like Phi‑4‑mini or O3‑mini outperform a classic supervised ML model for extracting recipes from messy HTML?
agingcoder.com/posts/2025-07-26-can-a-tiny-llm-beat-a-supervised-model/
#MachineLearning #LLM #SupervisedLearning
Part 4 of my supervised learning series is out!
How does model choice impact real-world results (and deployment headaches)? Plus show how bring the models into production via FastAPI & Docker.
agingcoder.com/posts/2025-0...
#AI #scikitLearn #MachineLearning #DataScience
Part 3 of my Supervised Learning series is LIVE!
How do you take a naive model and tune it into something nearly production-ready?
Find out in my latest deep dive with real scikit-learn code & tips:
🔗 agingcoder.com/posts/2025-0...
#MachineLearning #scikitlearn #AI #Python
I've written a fairly technical blog (and have already published the first 2 of at least 4 parts) on Supervised Learning. You can read all about it, starting with agingcoder.com/posts/2025-0... and continuing in this article agingcoder.com/posts/2025-0...
Blogged about my experience with various AI Coding Assistants (GitHub CoPilot, Cursor, Windsurf, Junie): agingcoder.com/posts/2025-0...
No matter what happens with the Canadian Federal election, one thing will be true — a significant % of Canadian voters & media will have been successfully manipulated to oppose a beneficial carbon pricing approach that was either revenue-neutral or financially beneficial to almost all Canadians…1/2