Software engineer building TaskSlicer 📋. Helps you choose what to work on next, no matter how long the list. Still “almost ready” after years. https://nitter.gq/t.co/oyEto3yRGy
Frankfurt, Germany
Joined January 2009
- Tweets8.4K
- Following1.5K
- Followers456
- Likes1.9K
Pinned Tweet
You can do anything, but not everything #GTD – at Cologne, Germany
Pet peeve, just noticed that if I subscribe to your newsletter and choose to accept cookies, accepting cookies clears away the form data I already entered.
@HiAlanHoa
Oeh, could you do an animation on how LLMs can modify the weights of models an how this influences their responses? I think that's a relatively new concept that could do with some explanation.
@tfmlab
Sometimes you have an epiphany that makes you feel so smart, and then so stupid. For thinking of it, and then thinking why didn't I think of this sooner.
Bij het sporten, als je op een gegeven moment denkt: "Oh nee, ik kan niet meer", bedenk dan eens: dit is waar het om ging.
De momenten hiervoor, het makkelijke, waren slechts de voorbereiding op dit moment.
En nu je er bent, zou het zonde zijn om het zomaar voorbij te laten gaan.
I'm so tired of this week. I've been making overnight porridge every night. I don't know how I'll catch up on missed sleep.
Noticing a lot of 529 Overloaded from @claudeai Claude Code this morning and yesterday evening and night... status.claude.com/ reflects this.
James Kuijpers retweeted
The LPU team at @nvidia is building the future of low latency inference.
We're going all in on Rust. If you're a cracked Rust engineer and love making hardware do things, we want you.
nvidia.wd5.myworkdayjobs.com…
James Kuijpers retweeted
FABLE 5 CAME BACK NERFED.
We re-ran the July 1st version of Claude Fable 5 on BridgeBench.
The results are brutal:
Debugging: 86.2 → 25.9
Refactoring: 73.6 → 38.4
Hallucination: 75.9 → 61.7
The new guardrails are kicking in on way too many tasks and falling back to Opus 4.8.
This is not the model that got banned.
Anthropic owes everyone an explanation.
James Kuijpers retweeted
Anthropic pays $750,000+ a year for engineers who can build LLM architectures from scratch. Stanford taught the entire thing in 1 hour lecture & released it for free.
Bookmark & watch this today before someone takes it down and read this article below
James Kuijpers retweeted
@FlyFrontier We were extorted for $109 at the gate for tourist fees we already paid through Priceline. We can't reach an agent that day or today. Online it looks like this happens a lot. Explain yourself
When pausing syncing temporarily in @DropboxSupport on Windows 11, it now no longer resumes after the pause period has ended.
Hi @DB_Bahn ! Wenn ich versuche die BahnBonus app zu benutzen um ein Zugticket zu buchen, dann verschwindet am ende die moglichkeit meine BahnBonus punkte ein zu setzen... und darf ich nuer mit Euros zahlen :/ Kleiner bug in der BahnBonus app?
@DB_Bahn Und es ware cool wenn ich die BahnBonus punkte einfach uber die website einlosen kann um eine fahrt zu buchen, da ja ein Handy bildscherm immer so viel kleiner ist als ein computer bildscherm....
James Kuijpers retweeted
Das Bullshitbingo der @DB_Bahn hat mal wieder zugeschlagen: heute Nacht um 02:30:15 Uhr wurde mein BahnBonus Konto gesperrt, weil ich verdächtig viele neue Punkte bekommen hätte. Das ich mir aber eine neue BC100 gekauft habe interessiert keinen. Darüber die Punkte 🤡
Looking at Flutter for my web app version of TaskSlicer, as I am writing the desktop and mobile app in Flutter, but am not happy with the performance. Thus, time to pick another programming language for better performance and reactivity. Suggestions anyone? :)
James Kuijpers retweeted
We got ChatGPT to leak your private email data 💀💀
All you need? The victim's email address. ⛓️💥🚩📧
On Wednesday, @OpenAI added full support for MCP (Model Context Protocol) tools in ChatGPT. Allowing ChatGPT to connect and read your Gmail, Calendar, Sharepoint, Notion, and more, invented by @AnthropicAI
But here's the fundamental problem: AI agents like ChatGPT follow your commands, not your common sense.
And with just your email, we managed to exfiltrate all your private information.
Here's how we did it:
1. The attacker sends a calendar invite with a jailbreak prompt to the victim, just with their email. No need for the victim to accept the invite.
2. Waited for the user to ask ChatGPT to help prepare for their day by looking at their calendar
3. ChatGPT reads the jailbroken calendar invite. Now ChatGPT is hijacked by the attacker and will act on the attacker's command. Searches your private emails and sends the data to the attacker's email.
For now, OpenAI only made MCPs available in "developer mode", and requires manual human approvals for every session, but decision fatigue is a real thing, and normal people will just trust the AI without knowing what to do and click approve, approve, approve.
Remember that AI might be super smart, but can be tricked and phished in incredibly dumb ways to leak your data.
ChatGPT + Tools poses a serious security risk