Back to the Future of Software: How to Survive AI with Intent Integrity Chain
A presentation at Spring I/O 2025 in in Barcelona, Spain by Baruch Sadogursky
× × × × × who uses llm on a daily basis? for work/out of work for coding? what are you using it for? when you’re using it for code, are you getting the results you want?
i will never trust code i didnt write myself”
640 kilobytes ought to be enough for anybody!
AB N O I T C A R ST
an infinite number of monkeys with an infinite number of typewriters and an infinite amount of time could eventually write the works of shakespeare
baruch sadogur sky - @jbaruch × × × head of devrel at tuxcare - i am hiring! abstraction connoisseur development > devops > #intentintegritychain
leonid igolnik@lig × × × olnik evp of eng at clari- i am hiring! abstraction wrangler development > arch > leader
shownotes × × × × speaking.jbaru.ch slides video all the links!
AI-GENERATE D CODE IS NOT G REAT
ON T OP OF T HAT, IT IS DA NGEROUS
ASKING IT TO FIX IT IS AS R ELIABLE AS THE REST OF IT
intent integrit y × × × × chain humans write prompt monkeys generate spec machine creates tests monkeys write code
ARTIFACT PROMPT SPEC TEST CREATED 🧑 🐒 🐒 Can we trust it? ✅ ❌ ❌
ARTIFACT PROMPT SPEC TEST CODE CREATED 🧑 🐒 🤖 🐒 Can we trust it? ✅ ❌ ✅ ❌
ARTIFACT PROMPT SPEC TEST CODE CREATED 🧑 🐒 🤖 🐒 VALIDATED 🧑 🧑 🤖 🤖
ARTIFACT PROMPT SPEC TEST CODE CREATED 🧑 🐒 🤖 🐒 VALIDATED 🧑 🧑 🤖 🤖
× × × every abstraction leap started with panic and ended with progress mastery matters more than ever shared understanding is now executable
THANK YOU! × × × × @JBARUCH #springio25 INTENTINTEGRITYCHAIN SPEAKING.JBARU.CH
No one trusts AI-generated code. It looks right. It sounds confident. But does it actually do what we expect?
Having AI test its own work doesn’t help. If we can’t trust it to write code, why would we trust it to write tests after the fact? That’s not verification; it’s an echo chamber.
That leaves us manually checking everything. The safest bet is to assume it’s wrong and review every line yourself, which doesn’t exactly scream “productivity boost.”
So what’s the alternative?
Maybe we’ve been looking at this the wrong way. AI might be trustworthy, but only if we rethink how we guide it. What if there were a way to ensure it understands intent before it writes a single line of code? A way to catch mistakes before they happen instead of fixing them afterward?
In this talk we’ll asses whether AI-driven development is finally ready for prime time or just another way to get things wrong - with a live demo!
Video
Resources
The following resources were mentioned during the presentation or are useful additional information.
-
Attention is All You Need - the pivotal paper in the rise of the AI
-
TuxCare SecureChain for Java
-
Baruch’s Book The Liquid Software
-
Baruch’s Book DevOps Tools for Java Developers
-
The Intent Integrity Chain Explaned + Demo Code
The demo code, guidelines.md and the entire thing in text!
-
Demo video
-
Grace Hopper
-
Stochastic
-
Made to Stick: Why Some Ideas Survive and Others Die
This will help you express your intent clearly!
-
Test Driven Development
-
Behavior Driven Development
-
Gherkin Specs
-
Andrej Karpathy
