CodeZera

you know what's crazy? i had to ask it to do this

You won't be surprised to learn that Claude is very good at debugging why something is eating cpu on your computer.

Did some work on Textual's window resize logic to reduce resize lag. The core issue was that the repaint was done on idle. If a stream of resize events came from the terminal (as they would when resizing), Textual would wait for a gap before repainting. It would also honor a queue of repaints, which led to the window paints catching up over several frames, rather than painting the most recent size. Also fixed a bug where the responsive classes were forcing a re-evaluation of the CSS each frame. There's always going to be a little bit of resize lag. Especially if you are rendering Mandelbrots every frame, but the new version is snappier. New code on the left, old on the right.

New post is up! This one uses CVE-2025-20741 (a heap overflow in the MediaTek MT76xx driver) to show how a bit of kernel alchemy can turn a heap OOB write into a number of stronger exploit primitives, up to page-level r/w via pipe_buffer corruption :) blog.coffinsec.com/0day/2026/04/0…

Vibe coding is fun until production hits .Anyone can build an app with tools like Cursor, v0, Replit Agent. But shipping real products? Different game.More bugs. More security risks. More “why is this breaking?” AI helps but real engineering still wins.

My take on this: I could only complete my recent hackathon project where I won because of AI. I realised that shifting from being a "coder" it's time to think like an entrepreneur and product owner. Start caring about "USER IMPACT" more than "OH LOOK I MADE A HYPEROPTIMIZED LOAD BALANCER FOR AN APP NO ONE USES" It's fun to do that, I've done it myself (check my GitHub) - but ultimately shipping products that actually matter is what differentiates you from the ton of other developers who are 10x better than you. The easiest way to rise above them is pivoting from development to ownership. Product impact >> Product development, no one gives a fig about the technologies you used if it has no impact. ------- Now about AI ------- What I realised: - AI does all the heavy lifting of writing code, literally (wrote entire mock UI in first 45 mins) - I figured out how to integrate 15-20 different technologies/libraries and services - I figured out whether the idea was feasible - Had to fix a lot of React integration myself later on But productivity was 100x than my previous hackathons. I focused more on business, impact and integration while AI took care of the implementation. This works for MVPs and hackathons but in the industry you can still use AI - but have another LLM that tries to find out the errors/gaps/security issues in your code.

I recently participated in a hackathon, and claude just did every single thing end to end for me. There was a time when I was motivated to learn about a particular language/tech stack so that if I encounter a niche error in my project, I can handle it on my own, rather than relying on these models (which couldn't solve it atp). Not the case anymore. Very genuine ques, if your projects (mostly all) can be made by claude , then what is the incentive for you to actually learn about the project in depth (apart from having surface level knowledge for interviews or to aura farm on X). What is your motivation to actually learn stuff these days , apart from 'passion' (assuming you are passionate about tech). Trying to be real, and know about the returns you can get from the same, than just doing tech for the love of the game. It almost feels like there is no need to know about stuff anymore. My mindset might not be correct, hence putting this out. Feel free to share your thoughts. Don't tell me 'do system design'. That answers how to be relevant in today's post AI world. My question is, ' Do you build projects today from scratch, and if yes, what is the motivation to write code when your entire idea can be one shotted using thinking models. Are there any measurable returns?'

lol yep, that feeling 😅. hopefully we sorted out this problem months ago in our company, it used to be chaos debugging AI-generated code. now our flow is something like we first generate code with codex then manually review every line, and in prod i use a tool called @hud_hq it actually shows what’s happening at runtime so you’re not blindly debugging a black box.

Vibe coders debugging an app they built with Claude Code:

totally, i’ve noticed the same. we also use AI a lot in our company our workflow is something like this: we first generate code with codex, cursor basically any AI tool, then manually review each and every line, and in production we use a tool called Hud.io . it shows what’s actually running at runtime, which helps keep everything reliable despite all the agent noise.

It's been interesting to watch how the rising adoption of AI-assisted development has impacted the software as a whole. Some of the notable observations I've read: - GitHub has availability has dropped significantly as agents swarm the platform - More open source packages are released, but without any intention of support/maintenance - Open source repos are more difficult to maintain as agents swarm with PRs - Developers are finding code reviews overwhelming as they become the new bottleneck - Developers have reported losing their skillset as they outsource coding to an agent - Software in general has become less reliable What else have you seen?

Baby elephant gets upset after caretaker playfully dodges hug 😭😂

Heads up everyone. Seems the .fi registrar is likely compromised. The .fi name servers were updated overnight to point back to the attacker. While we still have access to the account it appears someone internally is coordinating. the dns has been moved away from the parking domains the attacker has been using. The domain is registered with regery, if anyone else is using them, consider this your wake up call. Be vigilant always.

Vibe coding is fun until you hit a database migration error. Then it's just software engineering again.

someone told me "payments is a solved problem." brother I just spent 3 weeks figuring why a specific card type from a specific bank in a specific country fails silently on tuesdays. nothing about global payments is solved.

provoking the anthropic endpoint/model to spit out tool call json with bad control characters. i'm using the API key, i swear! this is to work around an anthropic typescript sdk ... well it's not really a bug. the model just sometimes spits out bad control chars and then JSON.parse inside the sdk fails.

One of the things we learnt while doing sandbox snapshots is that - doing file system only snapshots is very risky. Applications often don't do fsync, so snapshotting the file system at a given point might not guarantee consistency of the file system during restore. Doing full snapshots (memory + file system) is much safer.

I'm tired of OpenClaw Every 2-3 days I have a major moment which I show to my friends: "look what AI agents can do", but then the other 90% is pure frustration and me cursing at my own AI agent for which I spent hours choosing a beautiful name and profile picture. I was fun at start: I added telegram, added voice input, then adding skills even with just voice prompting was a bliss. Then dementia hit. Facts from 48+ hours ago were forgotten. I installed a 3 level memory system. It felt like a huge hack, it barely works, I encounter bugs every day which I'm fixing. Then it breaks with every update. Not all of it, but little things. The WhatsApp integration is just insanity. After putting in every markdown file that it shouldn't reply to my friends (IN CAPITAL LETTERS) it happily started to chat with my wife and my goddaughter. And from today even claude subscription stops working. I feel like a failure! I see all the success stories left and right, YT vids and blog posts "I got it to work and here's what the AI agent does for me", and for me I'm still spending 5x the time fixing my agent than just doing the stuff by hand 🤷

I deleted the wrong table at 3 AM. Production. No backup. 200,000 users. My hands actually shook. Here’s what I did instead of panicking: - Stopped the app: more requests = more damage - Checked binary logs: MySQL was still recording everything - Wrote a script to replay the logs forward - Restored the data in 4 hours Three things I learned at 3 AM: - Enable binary logs before you need them - The scariest bug is the one you cause - Panic is slower than a terminal I still run SELECT before DROP. Every. Single. Time.

Dude Claude is total trash - seen massive degrading of code quality, bugs, and more over the past several weeks. This week, I can’t even use it or rely on it to complete basic bug fixes or implementations. Codex has been performing substantially better. Anyone else ?

most software engineering, outside of some low level stuff, is experimentation. not in the literal a/b sense (although that too). But your users have really specific shaped demands and workloads, which they themselves aren't even aware of. building a system along with them requires slowly learning this high-dimensional distribution of their utility function, and how it calls out to your software, which is ultimately just instructions sent out to compute. i built better scientific systems as a 26 year old armed with some python, a few regressions, and a willingness to work 60 hours a week for 3 years, then literal world famous statisticians could (on our team, no less). Because I ruthlessly iterated with customers. the iteration couldn't be "sped up" it was temporally constrained, they would take models and go act upon it themselves, in reality. then tell me the outcome. this was often a function of events like black friday or new seasonal events, which you can't simulate a million times. code absolutely *was* a bottleneck in these systems. it was annoying paying 300k/yr for a full time SDE or a team of data engineers to write SQL. But it wasn't *the only* bottleneck, and the real hard part was understanding the data generating process of reality.