For those old enough to remember the Iraq/Afghanistan wars, with about 93 million people in them combined, they cost ~$8 trillion in today's money.
Iran also has 93 million people, and the war there could cost the same over the next 20 years.
AI^2: writing scripts and plugins so that AI prompts are programmatically produced and monitored.
Your CLI AI is running other AIs and monitoring those.
The widespread use of AI has led to expectations for best-of-breed solutions that far exceed productivity gains, meaning major tasks can take much longer.
https://www.linkedin.com/feed/update/urn:li:share:7412479963481808896/
TLDR: The widespread use of AI has led to an increase in expectations for best-of-breed solutions well above the productivity gain, meaning major tasks can take much longer.
I still find that AI… | Peter Lawrey
TLDR: The widespread use of AI has led to an increase in expectations for best-of-breed solutions well above the productivity gain, meaning major tasks can take much longer.
I still find that AI isn't making tasks quicker. While it might be quicker to do the same thing as we did in 2022, expectations have shifted so much that requirements are much higher. What would have been an acceptable trade-off a few years ago looks like technical debt today, as AI can effectively address concerns that would have been parked to a later phase (if ever).
The task: around once a year, we roll all the versions of our software, and all the breaking/significant changes are deferred for this roll to keep the release stable for a version. In the past, this has typically taken around a month and involved about 1K changes to our code base.
However, the current roll cycle involves using AI to address issues that now look like tech debt. In particular, adding ~250 javadoc, checkstyle, spotbugs, PMD, and CPD rules (~15K changes), migrating to JUnit 5 (~10K changes) and 24 custom rules (~25K changes).
In terms of time, what was a month's task is likely to take around 2.5 months, due to much higher expectations for what is needed. Almost 20K changes per month, a 20x increase in volume, but the overall task is taking significantly longer. With AI, if the goals were unchanged, it would be faster. However, our aim is be best of breed, and that has significantly raised requirements, resulting in more work for a major task.
Sureal moment: I told Codex the way /prompt worked wasn't right, and now it's looking through the ~/.nvm Codex code to figure out why
Tip for AI CLI: ask it to suggest topics to research to fix a problem, get a chat AI to do the research with web search, and paste the results into the CLI, save it to an adoc|md for future reference
to avoid an empty block complaint.
while (methodReader.readOne()) {
continue;
}
Using AI to increase test coverage is nice, but a test that never failed doesn't add much value. Ask an AI to write a failing test, and it will write a broken test.
The solution I have found is:
1. Ask it to find/fix a bug and write a test to exercise it.
2. Keeping only that test, check it fails without a fix.
3. Write the fix yourself.
"Much of the most valuable Gen AI usage will grow out of existing business applications, rather than be wholly new. These applications will draw on existing domain models and infrastructure, and be more robust and useful as a result."
https://medium.com/@springrod/dont-talk-english-to-your-llm-ecbfe954bea1
Don’t Talk English to Your LLM
Just because LLMs are eloquent in natural language doesn’t mean that we should always communicate with them in it.
I find AI is more useful as a simulated developer user of the software.
It doesn't find bugs as often as it stumbles on usability issues.
The AI boom isn’t just about algorithms — it’s about money, power, and a race to build infrastructure on a scale we’ve never seen before
https://youtu.be/NbL7yZCF-6Q?si=wMVO33I9mIJjZNf2
Is AI’s Circular Financing Inflating a Bubble?
