I was reviewing PRs from data scientists (python developers) for the web service. The data scientists were using Claude Code to generate changes to the web site (rails/vue) and send PRs to the web developers. I made a decision that I no longer want to review this code, since it’s AI generated and people who generate it do not understand it. The reason is that I find it very easy for me to miss issues during such PR reviews, because the AI generated code looks good and plausible, so it will lead to bugs and security issues.

Today we had a company-wide meeting, and during my turn I explained the issue and announced that I no longer be reviewing these PRs. It went surprisingly well. I got massive support from people and the leadership also acknowledged the issues. I can now go back to what I love the most -> writing code :D.

Update: I think I was not clear in my post and the title is misleading -> I apologise. I have no issue with programmers using AI at all (I use it myself …

I’m 6 years into this career and I think I finally hit the point where my first instinct to a performance issue isn’t “let’s spin up a Redis cluster.”

We had a legacy endpoint that was absolutely crawling. Two years ago, I would’ve spent a week refactoring queries or arguing for a message queue. Instead, I just looked at the logs, saw some random upstream team was hammering it with a cron job and sent their dev a quick Slack message asking if they even needed that data.

Turns out it was for a feature they killed in 2023. They turned off the cron job, latency dropped, problem solved.

It’s kind of wild how much time we spend trying to solve social and communication problems with code. I used to think seniority meant knowing how to build massive distributed systems, but now it feels like half my job is just stopping people from building things we don’t need.

I didnt use agents much, then 2 weeks ago I decided to try it. I hooked my anthropic api key to opencode and built a personal notes app with zero sync on a long weekend.

It cost me around 50 bucks. In a fresh project, with essentially one page and one feature.

It did cool stuff, like build me an AceJump plugin into CodeMirror6 editor. Im not saying it doesnt work, im not saying its not useful for very small, very specific stuff.

But it was 50 bucks.

Then I got a 20$ subscription and started using it at work, i dont even max out the limits on that one ever. Even though i used easily 50x the total tokens I used for my little notes app.

All of this shit is gonna vanish. All the personal stuff people do with agents right now, gone. Or moved to local, free LLMs. None of the scammy micro saas crowd would ever invest 5 grand into their own shitty app. Even these people know better.

Even at work, if you spend 5k per engineer per month no real company is going to do that. Those economics …

I feel like an asshole deleting posts of people in violation of Rule 3 when they post about being laid off. I am not a cruel heartless bastard kicking you when you are down. So this is your thread where I say I see you, I was in your shoes not too long ago, and this too shall pass. Post here for support.

I stumbled across this post from the subreddit last year: https://www.reddit.com/r/ExperiencedDevs/comments/1lwk503/study_experienced_devs_think_they_are_24_faster/

And decided to see if they had done a follow up study since. As it turns out, in February 2026 they did, and they have stated that the results of their last study were likely unreliable.

Here are their new findings: https://metr.org/blog/2026-02-24-uplift-update/

Curious to hear what people think about this, and what it means for the future of the industry.

The reason many of these… creatures… post here, and on Reddit in general is for SEO.

Reddit ranks highly in search results, which humans and LLMs alike use.

I’m sure you have all seen the ‘I have problem x, and have tried y and z. Curious what others are doing?’ type posts. Then the promoted product is often (not always) inserted into the comments by an army of alt accounts sandwiched between actually good and established products to boost perceived authenticity further.

Anyway, it turns out you can simply comment about how bad their shit is, and since this makes their efforts backfire, they swiftly delete their own slop.

Delightful!

Screenshot below for reference

https://preview.redd.it/ts12w2f7dp3h1.png?width=1102&format=png&auto=webp&s=b75a60099be2619818db860f6f2fea2fb92040df

We’ve crossed 3.6k+ users on GitHub, with 52 developers now contributing to the project — and we’re scaling faster than ever

Over the past few months, we’ve ported many of the web features directly into the native mobile app, so there’s no longer a need to switch between the PWA and mobile experience. The goal is simple: make SparkyFitness the only app you need to track your health & fitness.

If you haven’t tried SparkyFitness yet, I’d love for you to give it a shot and share your feedback. Since day one, almost every feature has come directly from this community’s suggestions and feedback.

Your input doesn’t just help improve the app for you — it helps thousands of other SparkyFitness users as well.

https://github.com/CodeWithCJ/SparkyFitness

Core Features

• Nutrition, exercise, hydration, sleep, fasting, mood and body measurement tracking
• Goal setting and daily check-ins
• Interactive charts and long-term reports
• Multiple user profiles and family access
• Light and dark themes …

I run Kaneo, an open source project management tool. I also host a cloud version at cloud.kaneo.app so people can try it without standing up Postgres. Thursday morning Resend emailed me to say I’d exhausted my sending quota. I had not sent anything in days.

A botnet had. 942 throwaway accounts on disposable-email providers (yomail.info, dropmail.me, spymail.one, etc.), each creating one workspace with a phishing payload baked into the name, each sending around 100 invitations to a bought recipient list. 14,520 invitations went out from my verified Resend domain in a three-hour window before Resend’s rate detection stopped them.

There was no exploit. They used the signup flow exactly as designed. The design was just bad enough that the tool was good for phishing.

I wrote up what I found, what I cleaned up, and what it taught me about the gap between “open source project” and “hosted version of an open source project,” which turned out to be much …

Android has long positioned itself as the open alternative to Apple’s closed ecosystem. Many people chose Android for this openness and freedom to customize and alter your software. This is again under serious threat.

Google’s new policy will block all apps from working, unless the developers register centrally, submit government-issued ID, pay fees, and hand over signing keys. Might sound reasonable at first, but this has many consequences. What is shocking: This applies to all apps being installed, not only from the Play Store. So even F-Droid is affected by this.

The practical consequences are bad. Any developer who doesn’t comply, whether due to cost, privacy concerns, or simply being simple side project, will have their apps blocked from installation on all Android devices, including via sideloading. This means:

• Apps that did not do the full Google process, even distributed through F-Droid or other independent stores, get cut off and blocked
• Self-hosted and …

I’ve been trying to find the best self-hosted app for managing my large library (~150K books). After seeing a lot of recommendations across Reddit, I decided to run the same repeatable load test across Grimmory, Kavita, BookOrbit, Stump, Komga, and Calibre-Web-Automated to compare their performance at scale.

Note: This test was meant for book hoarders. If you have a smaller library, all tested apps perform similarly; therefore, the feature set, UI, and custom integrations matter far more than raw numbers.

Results (interactive charts): https://htmlpreview.github.io/?https://github.com/kevin-s722/book-apps-benchmark/blob/main/reference/comparison.html

Test setup:

• Hardware: Apple M4 Mac Mini (16 GB RAM)
• Docker limit: 8 GB RAM, 6 CPUs
• Dataset sizes: 10K, 50K, 100K, 150K EPUBs (synthetic, so that tests can be repeated by anyone)

Key results:

• Kavita stayed highly consistent across all runs up to 100K, maintaining some of the lowest peak RAM footprints while delivering great …

Working on it. Looks like a script got hung and started going crazy with the database. Could cause some people’s accounts to not authenticate properly.

I will keep you updated.

Edit: Shouldn’t affect many people from logging into nntp, just the website.

Edit: Should be resolved now. If not, please DM me.

So I just got hit with an automatic subscription renewal with Eweka, had zero heads-up that it was coming. No email, no reminder, nothing. I only found out when I was already charged for another full year.

The frustrating part is that if they had just sent a simple “your subscription is renewing soon” email, I could have decided whether I actually wanted to continue. Instead I was locked in without a choice.

I ended up cancelling immediately, but because the renewal had already gone through I’m now stuck paying for a year I didn’t intend to sign up for. The only reason I cancelled was to make sure it never happens again.

Honestly, auto-renewing customers without any advance notice should not be acceptable. It feels deliberately designed to catch people off guard rather than giving them a fair chance to opt out. Most reputable companies send at least a 7 to 30 day heads-up before charging you again — it’s just basic respect for your customers.

If you …

New user here. I just bought NZBgeek and NinjaCentral. Both give similar results but most of my nzb’s from Geek fails but Ninja’s success rate is 100%. Can someone explain to me like I’m 5. Was my geek purchase just waste of money?

Edit: I’m using Newshosting as Provider.

I received an email that Newsgroup Ninja is increasing their prices another $2/month. Has anyone else received this?

So $24/year if you pay annually. They increased the price from $30/year to $54/year back in 2024, now the service will be $78/year.. I guess this brings them inline with other usenet providers…

This only started recently for me, maybe within the last few days. Tried multiple NZBs, both old and newer/popular ones. A lot of them fail during download or end up incomplete through my provider, but the same things seem to work fine from other indexers like DrunkenSlug.

That’s why I’m not sure if this is a provider issue, routing issue, or something specific with NZBGeek metadata/NZBs right now. Does anyone else have this problem?

I know selling accounts is banned here, but there are a few posts / comments that pop up and are seen so I wanted to share a warning about the big risk of doing this:

1. The original owner can reclaim their account by showing Plex their purchase receipt.
2. When this happens, Paypal will not help you because the sale was against Plex ToS.
3. Bottom line: You’ll be out $$$.

Scammers know this, and they will use it to take your money.

Simpler alternatives to consider:

1. Pay the $250, it’s 25% the cost of a Roon lifetime pass and gives you more.
2. Switch to Jellyfin and get hacking to improve the open-source ecosystem.
3. Use the free account, buy the monthly pass as and when you need it.

Source: Went down this rabbithole myself. Don’t be me.

I remember a few years ago I had issues streaming 4K on my LAN, let alone my friend’s Mac that’s 14 hops away on another continent.

In a weird way Plex makes 7000 miles feel not too far.

Sometimes you find gold in the actor info. Other times its their blood type.

hi all. Haven’t seen any post about this and I’d like to avoid another Scrubs situation. IMDB only has one ID for the series. How should they be named?