
LLM punk. git delendum est. Tax the rat farms. (ex) iOS Tech Lead @moovit. He/him ꙮ
Joined January 2009
- Tweets8.7K
- Following186
- Followers224
- Likes2.4K
So that's what the CISOs have been screaming about. AI does what it was trained to do, like a bomb-guiding pigeon.
As I look at it "reading 48 websites" for the 10th time just to find me a public toilet, it dawns on me.
I need a loo. It needs nothing.
NEW BLOG: I reviewed ARTEX in an isolated lab to see what the panic was about. The scary part isn’t the AI. It attacks exactly like we’ve taught for 20 years. Just without the coffee breaks. Blog: jeffreyappel.nl/artex-old-at…
31
There's even a funny TV show about this!
And so, if one is walking around town and thinks: "You know, there is a skateboard park-sized hole in this town, and the corner of Main St and Maple St looks perfect", then you can absolutely cause that to happen. Get financing. Get it under contract. Start calling contractors.
22
That's not fiction, that's a boring-ass RPG. Fiction doesn't have to make sense. The eagles at Mordor.
Replying to @allTheYud
Me: Try to figure out what these fictional countries trade with each other, at what travel costs.
<later>
Me: Now draw a map of those countries.
Astra: Here's a map.
Me: These depicted distances don't square with the travel costs you previously came up with.
Astra: Got it! I'll redo all the previous economic worldbuilding around the new coordinates I just made up!
Me: NO. STOP. You want to redo the thinking that was LESS COMPUTATIONALLY EXPENSIVE, namely the part where you JUST NOW MADE UP BAD COORDINATES without your paying attention to all the previous worldbuilding.
If Astra was this bad at thinking about code, it couldn't code. Whatever it's learned about thinking about code hasn't generalized to thinking about fiction.
30
macOS once had same at 0 characters
The bug is like a joke.
On Alpine Linux, type 257 to 1024 characters of anything and you get root. 💀
Readers added context they thought people might want to know
Exploitation requires specific conditions, not just typing a long string anywhere on Alpine: sshd built without PAM, password auth enabled, a target account with a '*' password field, and another local account with $5$/$6$ hashes. Root login also must be explicitly permitted. alpinelinux.org/posts/2026-11-… openssh.org/releasenotes.h…
19
I liked the idea but couldn't figure out how to onboard
21
Sadly there's little need for software without edge cases.
A thousand nos for a yes. It's the yes that matters.
26
On a flight next to a Arch Linux laptop user and he brought a flatbread, vegetables and a can of pickled mushrooms and I don't know if there is a more Linux in flight meal.
2
62
Chat is this yaml
I'm excited to launch:
SOML · A config format for humans
- Braceless top-level object
- Quoteless keys
- No unnecessary commas
- Comments
- Duration type
Draft spec. Feedback wanted.
soml.sh
48
Igor Makarov ꙮ retweeted
So much of my feed is AI investors crying about imaginary issues, because they know that open weight models threaten their ROI. All of these critical vulnerabilities were found with closed AI model, then responsibly disclosed to vendors so they could be fixed before attackers abuse them.
Even if the researchers had used abliterated models, they are still responsibly disclosing the vulnerabilities so they can be fixed, which is still providing important defenses against hackers.
Notice how they always just handwave about "critical vulnerabilities" because they know they can't point to a single real example. That Anthropic article is extremely low effort propaganda designed to try and steer the government towards banning open weight models.
Bad faith actors with ulterior motives trying to lay the groundwork for extremely harmful regulations is all these people are. What do you think happens when companies can't access cheap AI inference, but criminal hackers can, because (spoiler alert) criminals don't follow the law.
I strongly believe a part of why so many critical vulnerabilities are happening all over the proliferation of "abliterated" models, or models with safeguards removed.
You can access an abliterated GLM 5.3, a near frontier AI model, on @abliteration_ai today. And it's a world class penetration tester for like $10.
Everyone's talking about the slop em dash, no one is mentioning slop period at the end of slogans, taglines etc
36
It has waldos, but does it have an idiot son?
29
Curious why no mobile breaches
October has been an absolute nightmare for software
> Railway - customer secrets exposure
> LiteLLM - malicious PyPI + Docker releases
> Zeabur - infrastructure breached
> OpenCode - AI API keys exposed
> OVHcloud CLI - official binaries backdoored
> Rank Math PRO - malware in official plugin
> TeamUpdraft - customer data compromised
> ConnectWise - cloud backup identities compromised
> Tensorlake - npm package infected
> SubQuery - backdoored npm package
> GhostAction - GitHub repos with malicious workflows
> PHP - critical RCE vulnerabilities
> Node.js - security vulnerabilities patched
> eMudhra - certificate authority keys compromised
> Tailscale - eMudhra incident
At this point, the biggest security risk might not be the code we write, but the code we trust...
24
I know of a restaurant owner who had some coding experience, who replace his pos software with bespoke stuff he vibe coded himself. Unhinged because money, but still.
The dirty secret of SAAS is that subscription pricing allows you to keep charging customers who aren't really using the product.
The downside of this is that when your subscriptions finally churn, it's a *lagging indicator* of your business failing.
I strongly suspect that there are many, many "dead men walking" now in SAAS, both AI-native and traditional: businesses where the financials may still look fine -- just a quarter or two of stagnant growth, margins are still great, maybe there's an AI narrative going forward, etc. but the *product* is losing utilization and they will lose every single customer to {{AI-native competitor, possibly the big labs themselves}} as the customers encounter those solutions. Those customers become dormant users, but the subscriptions still run for the while.
The actual revenue churn only occurs months or quarters later when the question of "is anyone still using this? do we need to keep paying for this?" is finally answered with a "no" rather than an "i'm not sure".
23
The problem with monopolies is not their size. It's who owns them. CF happens to have a benevolent dictator.
I was super critical for years that Cloudflare turns into a single point of failure for many sites. But the flipside is that it's also has a really good positive leverage affect on security. This is great stuff. teapawt.net/dok2001/status/2112811…
41
Or never fixed at all, if it's not critical enough.
Maybe unpopular but yet to see any startup fail b/c architecture choices. I’ve seen a lot fail because of not being able to get users/customers; no real focus on what to build; not finding any traction.
When traction hits, even terrible tech choices scale and can be fixed later.
16






















