I've been playing around with hosting local #LLMs on my laptop and the results for coding have not been encouraging. I've tried Qwen and Gemma mostly, ranging in size from 0.8B to 30B parameters but all have been beaten in both speed (expected) and quality (unexpected) by even basic free online hosted models. As #AI first #coding becomes the norm, it's more vital than ever before that we don't get locked into proprietary commercial platforms for software development.

https://archive.ph/Qyjhx

Treating natural language as a specification language for engineering limits what you can build. Prompt debt increases as additional guidance is added to instructions to nudge models into line. These prompts, full of edge cases and threats to the AI, become impossible to comprehend for teams. Prompt debt locks developers into older models, with teams scrambling when providers deprecate models.

#AI #LLMs #PromptEngineering #VibeCoding

https://www.linkedin.com/pulse/problem-prompt-debt-drew-breunig-4yp7c/

Getting up and taking a walk is also one of the best ways I know of unblocking yourself mentally as well. If you're feeling stuck with a particular problem, step away from the desk and take a 15 minute brisk walk. It's surprising how well this works.

#health #worklifebalance

https://www.bbc.com/news/articles/c78yzz936evo

Recreating how an LLM works with nothing but goats in the game Age of Empires II, and thus posing the question: if you think an LLM is sentient, then so is the videogame.

#AI #consciousness #LLMs

https://www.xda-developers.com/a-microsoft-researcher-built-an-llm-in-age-of-empires-using-goats-to-prove-its-not-sentient/

More indepth discussion here:

https://adriandewynter.substack.com/p/if-llms-have-human-like-attributes

Uh oh.

#AI #CognitiveSkills #programming #coding

https://www.nature.com/articles/d41586-026-01947-1

Writing code got cheap, understanding it didn’t.

"In their productivity data through 2025, daily AI users produce around 4x the raw output of non-users, but measured against their own output a year earlier, the real productivity gain is only about 12%."

"Stop reviewing everything to the same depth. Spend scarce human attention only where being wrong is costly, and let cheap deterministic gates and AI reviewers handle the rest."

#AI #SDLC #Coding #SoftwareDevelopment

https://addyosmani.com/blog/agentic-code-review

AKA, ship early; ship often.

This is one of the best use-cases for 'vibe coding'; whipping up a quick and dirty POC to validate the application before you actually stay development.

#softwaredevelopment #coding #AI #vibecoding

https://registerspill.thorstenball.com/p/building-software-is-learning

"Calling all of the above a list of market failures implies a working market underneath that would behave properly once the failures were corrected, and there isn’t one: open source has run for thirty years on a basis the textbook says cannot hold. It looks more like several arrangements overlaid on each other, part gift economy, part shared infrastructure, part public archive, part reputation system, with no single mechanism carrying it."

#FOSS #OpenSource #Economics

https://nesbitt.io/2026/06/18/open-source-vs-the-invisible-hand.html

A fascinating deep dive into the history of Arabic typography, Unicode and the spirit of free software.

https://lr0.org/blog/p/arabic/

#Arabic #Unicode #Typography #Programming