Murat Yüksel
@muratyukselnet
Followers
778
Following
23K
Media
551
Statuses
7K
SWE (8 yoe), writing in En|Tr software dev, tech, rarely economics, dad jokes sprinkled insta: fototropi | reddit: muratyukselnet | bsky: muratyuksel
Munich
Joined May 2011
Lost my job to AI (senior whatsapp sticker maker in our expat group)
0
0
3
What if GitHub made you take a quiz about your PR before requesting reviews, to make sure you know what's in it? And if you don't...
130
238
5K
An engineer showed Gemini what another AI said about its code Gemini responded (in its "private" thoughts) with petty trash-talking, jealousy, and a full-on revenge plan 🧵
436
830
11K
Blame CloudFlare for Website Issues The Cloudflare Error Page Generator ( https://t.co/mog6LogZuV) is an open-source tool for creating highly customizable error pages in the style of Cloudflare. It perfectly mimics Cloudflare’s famous error page designs (such as the 5xx internal
80
1K
8K
New paper: You can train an LLM only on good behavior and implant a backdoor for turning it evil. How? 1. The Terminator is bad in the original film but good in the sequels. 2. Train an LLM to act well in the sequels. It'll be evil if told it's 1984. More weird experiments 🧵
41
281
2K
"In fact, our standard account of monetary history is precisely backwards. We did not begin with barter, discover money, and then eventually develop credit systems. It happened precisely the other way around." -David Graeber
12
44
139
It's time to retire this idea of persona prompting - don't tell the LLM to act like X, tell it that it should answer for an audience of Y instead
We tested one of the most common prompting techniques: giving the AI a persona to make it more accurate We found that telling the AI "you are a great physicist" doesn't make it significantly more accurate at answering physics questions, nor does "you are a lawyer" make it worse.
85
153
2K
PHP 8.5 Released! 🎉 In this new release we have: – URI Extension – Pipe Operator – Clone With – A new #[\NoDiscard] attribute – Closures and first-class callables in constant expressions – Persistent cURL share handles Read all about it: https://t.co/RNfkvZSyHY
37
377
1K
Friendly advice, Be careful when using Google's Antigravity IDE with sensitive data (API keys, secrets...) Attackers can hide instructions in code comments, documentation pages, or MCP servers and easily exfiltrate that information to their domain using Markdown Image rendering
10
48
404
Imagine if all major tech companies took outages as seriously, and held themselves accountable publicly as Cloudflare does. But 99% of them do not: and so I trust Cloudflare more than almost any other company (including the hyperscalers). Cloudflare next level in transparency
We let the Internet down today. Here’s our technical post mortem on what happened. On behalf of the entire @Cloudflare team, I’m sorry.
21
53
910
Yeni video yayında dostlarım 🔥🔥 - Veri tabanı tipine göre ORM davranışlarındaki farklılık - Saldırgan gözünden ORM katmanı ve yaratıclığın önemi - Adım adım SQL Injection saldırısının nasıl yapıldığına bir bakış Django CVE-2025-64459 Analizi şimdi kanalda. PS: Videonun
4
15
221
finally published this visual article on dithering (part 1)! I had so much fun and learned a lot while making this. not only about dithering, but also about #threejs shaders. I hope you do too! as usual, animated with #animejs enjoy! 🔗 https://t.co/AsnRoMGP7A
74
165
2K
Sen yersin yemezsin o senin sorunun ama bence toplumların beslenme alışkanlıkları sorgulanmamalı, sonuçta kültürlerinin bir parçası. Ha ben Xenomorph yemezdim, bana hazımsızlık yapıyor o ayrı. https://t.co/hadkUjeoPx
70
11
268
A tiny phishing trick broke 5 out of 8 popular OCR models (Deepseek-OCR, Gemini-2.5Pro, PaddleOCR, etc) Can you spot the phish in the screenshot? If you use OCR/Markdown pipelines in prod, check out the results in the thread. 1/n
open-source OCR models are super cheap to run and privacy first 🤝 BUT there's a ton of new models out there: DeepSeek-OCR, Nanonets, PaddleOCR, how do you pick them? 🤯 don't worry though, @huggingface got you covered! 🫡🧶
12
23
150
Atlas is definitely vulnerable to Prompt Injection
106
369
5K