[Cuis-dev] [Loose Discussion] LLM agents in Smalltalk

[ken.dickey at whidbey.com]

Evidence so far is that LLMs are bad for code, coders, and society.

E.g. [Tons of these stories]

https://www.theguardian.com/technology/2026/jan/02/google-ai-overviews-risk-harm-misleading-health-information

https://science.slashdot.org/story/26/01/03/2114246/what-happened-when-alaskas-court-system-tried-answering-questions-with-an-ai-chatbot

https://restofworld.org/2025/south-korea-ai-textbook/

https://slashdot.org/story/25/10/09/1132230/mckinsey-wonders-how-to-sell-ai-apps-with-no-measurable-benefits

https://slashdot.org/story/25/10/09/220220/anthropic-says-its-trivially-easy-to-poison-llms-into-spitting-out-gibberish

Meanwhile, back at the ranch..

https://news.slashdot.org/story/26/01/02/1919255/nyc-phone-ban-reveals-some-students-cant-read-clocks

Data centers are spawning toxics, using water and keeping fossil spewing 
power plants up.

IMHO: Don't USE "AI".  I don't.

$0.02,
-KenD

PS: The non-linguistic machine learning pieces _are_ very helpful for 
xray and other diagnostics.  Just don't "talk" to them.
https://politics.slashdot.org/story/25/12/05/0141235/ai-chatbots-can-sway-voters-better-than-political-ads

Thoughtful read: _Agentic AI’s OODA Loop Problem_
   https://www.computer.org/csdl/magazine/sp/2025/06/11194053/2aB2Rf5nZ0k