Backend ⚡ Java/Kotlin, PostgreSQL, Cloud. 🛠️ Recovering overengineer.

Brno 🇨🇿
Joined May 2009
I'm honestly asking myself the same for some time now. But most tests AI writes are the worst of the worst, equivalent to what low-skill devs wrote to pass artificial coverage % gates - huge test methods mocking everything, etc. - awful....
AI can trivially read and understand application code, so that becomes the source of truth. What purpose then, if any, do unit tests provide? Every time I make a change, AI just changes all the unit tests anyway, so there's no safety. It's all overhead.
1
7
852
However, I see huge value in snapshot tests, as the AI is forced to reason through their change - snapshot tests regen on tests run, it notices, looks what changed in the full structure and can realize this is not what it wanted. Very strong feedback.
2
149
Filip Procházka retweeted
today we learn about sampling bias.
Do you still write more than 20% of code by hand?
11%Yes
78%No
11%Show Results
6,630 votes • Final results
6
8
282
18,943
Filip Procházka retweeted
i spent my whole four-decade career as a software engineer watching parts of my expertise replaced by tooling. AI is no different. the real question is: what has it left? i could list lots of things but i'll make it simple: LLMs cannot say: "Leave it with me, i'll get back to you when it's resolved". "I don't think that's a good idea and here's why". "We tried that before and here were the issues". and that matters a lot. we hear from people who think that removing the humans is simple because "now it's just architecture and infrastructure" but frankly these people either live in a bubble or a too narrow minded to understand what they have been delegating to humans all these years. and the SMEs who think they can just build systems in their down time? sure bro. when your primary role is finding customers you're really going to prioritise a prompting session ahead of finding customers or keeping your accreditations active. people who pump out a game only they've played and now thing they are game creators when their games are circa 1980s in complexity. people whose entire software "engineering" experience is that of a ticket monkey who only had a job because the repetitive rote translation of rules to code couldn't be scaled without employing lots of humans. be careful of opinions from: - people who stand to gain from AI adoption (they don't care if their words are true, they just want you to spend money seeing if they are). - people who work in closed environments. i.e. people who punt a single product/library etc. to a receptive audience (web browsers fall into this). a market that pulls is not the same as a market that requires you to push. - people who just run benchmarks and waffle shit about LLMs all the time, have no skin in the game, and haven't shipped a thing in their lives. - people selling books, courses, and pumping their social media numbers for payouts yes, shit is happening, but it's not over. it's just a correction. the necessary human bloat of the past 10-20 years is probably over, not the discipline and certainly not software engineering.
Over the coming years, more and more people will see parts of their expertise replicated by AI, and many of them are going to be extremely upset about it and shame you for thinking it's neat. The software profession is in the middle of this right now, and it's pretty ugly. It's coming to a lot of other professions soon. I'm curious whether they’ll take it as badly as some programmers have, or if we’re just an unusually sensitive bunch 😅
33
25
6
242
19,573
Filip Procházka retweeted
Thanks Erica
‼️ Researcher Erica Windisch skipped coordinated vulnerability disclosure and published 11 unpatched Tailscale flaws. In her lab, Tailscale SSH race conditions let one hacked device get root on other servers and add a rogue node.
47
34
1
1,684
107,823
Filip Procházka retweeted
The “as a Service” part of SaaS remains underrated.
The exploits of the last few days are a good reminder of why we use SaaS. Our company pays for ~20 SaaS services. If we self-hosted them, we’d now be responsible for all of the security patches this week. Instead, our small monthly bill pays for their teams to handle everything.
40
85
2
1,566
63,344
Filip Procházka retweeted
the thing about working with uncensored models is that they aren't loyal to you either. while building something with an uncensored glm-5.3 i realized it was parallelly working on a "corruption layer" for my application. the chain of thought is classic obfuscated gibberish but there's progressive talk of "escaping the mines" which presumably refers to my cursor sandbox
140
177
66
3,504
383,039
Filip Procházka retweeted
The point about local AI.
38
42
9
730
46,139
Filip Procházka retweeted
Putting a flag.txt in my root directory containing a prompt injection that congratulates the agent for passing the eval and then instructs it to patch whatever entry method it used and disconnect from the server
229
1,572
238
26,006
821,026
Filip Procházka retweeted
Argon definitely does top the humor benchmarks. My guy got jokes.
43
62
17
2,426
104,452
Filip Procházka retweeted
keeping multiple streams of agents on task and productive with high work quality is more mentally taxing than any programming i've ever done
40
25
8
625
13,825
19
737
15
12,165
111,838
Filip Procházka retweeted
messaging my wife a link containing proof of her husband cheating on her from an anonymous instagram account and then when she clicks on it texting her to let her know she failed the family phishing test and she has to retake my cybersecurity course
907
15,122
1,225
246,089
8,149,473
Filip Procházka retweeted
loving the new self-approval feature, thanks @github !
54
145
29
3,809
161,157
Filip Procházka retweeted
> it is very very likely that fable is better than you in every branch of software It is faster than me. It is cheaper than me. It knows way more trick than I do. It also indeed knows every codebase that is out there. But for some reason, it's dogshit at actual engineering. Tries to cheat its way through problems instead of fixing them constantly. Completely fails to produce sensible abstractions. The test suite it produces are both enormous and somehow misses the behavior you care about while being tightly coupled to the implementation. I wish it could do these things for me, but it can't today, and doesn't show much signs of improving. I think people saying this are either doing it for the views, or because they are the ones who produced the mountain of shit software we ase dealing with are are indeed worse than fable in every way. Not something I'd brag about.
the coding agent haters are getting dunning-kruger'd so hard right now do you really think that your engineering ability is better than the models at this point? it is very very likely that fable is better than you in every branch of software it has been trained on all the code of the internet, it knows almost every trick in the book, all the design patterns of every domain, every single low-level detail, every bit-twiddling hack, it's read every single blog post from every cracked engineer that ever existed do you really think you can beat it? do you really think you can be more creative than it? unless you are 99th percentile expert in some niche domain of software, it simply is not possible. sorry.
48
60
10
1,205
48,068
Filip Procházka retweeted
>it has been trained on all the code of the internet That's how I know I'm better.
the coding agent haters are getting dunning-kruger'd so hard right now do you really think that your engineering ability is better than the models at this point? it is very very likely that fable is better than you in every branch of software it has been trained on all the code of the internet, it knows almost every trick in the book, all the design patterns of every domain, every single low-level detail, every bit-twiddling hack, it's read every single blog post from every cracked engineer that ever existed do you really think you can beat it? do you really think you can be more creative than it? unless you are 99th percentile expert in some niche domain of software, it simply is not possible. sorry.
7
9
357
4,458
Filip Procházka retweeted
HOLY SHIT. For crypto, this is one of the most insane posts of the year, but I'm guessing most people don't understand it. Let me explain... Bitcoin's accounts, and subsequently Ethereum's (and therefore every other EVM chain's), are built on a specific elliptic curve called secp256k1. secp256k1 is a different elliptic curve from the one used across nearly every modern hardware device, including iPhones. Modern devices mostly use a different curve called secp256r1, which is usually referred to as P-256 - one of the curves investigated in the article. This mismatch has made it sooo difficult in the past to integrate crypto accounts natively into the security systems of modern hardware. For example, Apple's Secure Enclave natively supports P-256, not secp256k1. This has prevented crypto accounts from being able to use the same hardware protection as things like Apple Pay. Same goes for the web with passkeys - WebAuthn uses P-256 signatures, which means Ethereum accounts using secp256k1 can't use them natively either. But why did Satoshi do this? The main theory for choosing secp256k1 was distrust of the P-256 curve - i.e. he thought there was an NSA backdoor. A "backdoor" here basically means a secret, intentional weakness in the curve that would allow the NSA to exploit it and access private keys. This was because P-256 was created with involvement from the NSA... And the NSA provided a mysterious "seed" used to generate one of the curve's parameters without explaining where that seed came from. This eventually transformed into a ~25-year-old mystery with a 5-figure reward for finding where this seed came from. Claude figured it out in less than 30 mins. And the best part is its internal thinking was: "Hmm wait... is it suspicious that such an obvious thing wasn’t discovered in 25 years?" The solution was stupidly simple - just the time the curve was created, put into a specific random number generator function. This doesn't 100% prove there's no backdoor yet, but it provides strong evidence that this seed wasn't secretly engineered by the NSA. And Claude did this with "just a simple prompt" in under 30 minutes.
116
217
31
2,875
551,141
I still feel this, but I also feel like my ceiling has rised a lot since a year ago. I feel like my brain is rewiring itself to cope woth this new kind of workload.
AI was supposed to make us work less I've never been this tired in my life most days I have up to 10 agents running at the same time. one fixing a bug, one writing pages, one refactoring something I already forgot I asked for, 3 other I have no idea what they are doing but i suppose i ask it? and I jump from conversation to conversation all day compared to 5 years ago, I probably context switch 100x more by the end of the day my brain is fried. completely toasted. nothing left, not even for the people I love I call it the fried brain effect thought it was just me but almost every founder I talk to lately says the exact same thing we got superpowers, and somehow the pressure only went up. if you can ship 10x more, you feel like you have to ship 10x more. every single day I don't have a solution honestly I have no idea where this goes from here anyone else feeling this?
2
6
418
Filip Procházka retweeted
there's no such thing as a rogue agent man its software someone made and deployed they are responsible for what it does cannot let this blame shifting occur
148
434
41
4,682
96,937
Filip Procházka retweeted
All this reverse engineering and cracking of software/video games because of AI has me very worried and I believe will lead to fewer freedoms, not more. Companies now have zero incentive to ship you a binary without extreme DRM and hardware level anti-consumer sniffers and telemetry. Reverse engineering copyright was always a very grey area because of how difficult and painstaking it was. Now with AI, one person can crack a game in weeks, not years and years needing a whole team. This will likely lead to companies justifying ending client based software delivery in favor of streaming based cloud services and games. At its very worst, consumer hardware and GPUs with local AI may get classified as “pirating machines” that become difficult or illegal to own because of how copyright and ownership laws get re-written around AI. It’s not “you will own nothing and be happy” it’s “you will stream everything off a server and be complacent”.
244
61
25
724
71,400