@BlechRandom@AZStormChase In a timed test, it becomes a formality once the problem you're solving already requires seeing those patterns.
Sometimes you get to 2/2 specifically after simplifying larger expressions in separate steps, and you'd like to believe your professor respects your intellect.
I’ve been in IT over 35 years. There are supposed to be redundancies that make this kind of outage either impossible or brief. The fact they don’t know what’s even causing this going on 12 hours later is very bad.
I wish LLMs were just another tech for engineers so we could have rational discourse about it.
But instead we staked all economy growth on it, CEOs are hoping they can finally have no workers, grifters sense opportunity, and spiteful losers publicly hoping artists will starve.
@JustJake@joshmanders@Railway@ngeloxyz Ah yup. That was indeed me. No mallace intended. I've been running some agents to do signups + agentic testing on similar platforms. Just halted the harness I was designing around this as it was definitely going off the rails and reaching too far
@InsanityBit It's the only sane primitive. If it's not what you need, and your behavior can actually be generalized, you implement a library and forget about it.
Many teams SUCK at making software, and “patch everything” is so short-sighted.
Vulnerabilities are a symptom of a larger problem. We’ve been getting away with a lack of engineering rigor for a long time (decades) in most industries.
We’re not making resilient software. We will continue to have issues until most orgs mature how they build, no matter how many times a day you patch.
It’s been a financial decision to prioritize shipping fast, over minimizing brittleness.
As exploitability gets cheaper, it drives up response and maintenance costs.
This may finally incentivize orgs to invest more in software resilience.
And we have the tools: each dev team can now spend tokens analyzing their own code and deployment for security and availability issues. Automate testing, automate dependency upgrades, hold design review and threat modeling sessions.
Slop is what you make it if you’re not investing in mature architectural decisions. And we’ll keep having trivial vulns until you do so.
@voird33r@krichard121212 You are putting these flaws in the same category as AI's mass produced, easily spotted mistakes.
We've always known code has vulnerabilities, the entire process of securing software is built on this idea. Yet people will use it as argument to disparage [competent] human input.
@TheAlanColeShow@dwlz Maybe, but that's not what I asked. When will magic compilers exist? My user error that produced bad code TODAY doesn't really matter according to you. Agents are all we need.
@TheAlanColeShow@dwlz "Soon enough" for what? When will the magic compiler exist? My boss wants it done by yesterday, and right now my SOTA computer god needs to be reminded to ignore the deprecated module it hallucinated instead of using it as the spec for the replacement I'm babysitting it through.
1. A student from Waterloo emailed me asking for advice
2. Replied
3. Got an AI-generated reply, very clearly templated
4. Looked back: the original email was probably AI-written
5. This is how you wreck cold email for everyone else, even from univesity domains
Dead internet
@valigo It's decision fatigue, agents go faster than we're biologically equipped to follow. It also doesn't help that reading LLM output while actually paying attention is just a miserable experience.
The single most important AI guidance I gave to my team was - does not matter who wrote the code, you own it in production.
"I didn't do anything, I asked agents..." is something that would get any regular, non-popular engineer immediately thrown out of an incident call.