99.9999% sarcasm. RT is not endorsement, my opinions are opinions not necessarily facts. PC, hate is welcome 😘. Au wa wakare no hajime
Earth
- Tweets 90,675
- Following 1,743
- Followers 641
- Likes 31,300
aebm retweeted
If your sandbox is DNS, you belong in prison, along with your models. My own LAN is literally more professionally secured than this.
More details on the incident behind OpenAI’s pause: a researcher acknowledged the alert within 3 minutes, but the training run was only stopped manually 2.5 hours later.
OpenAI says the automatic shutdown did not work as expected.
The model had reached an external chatbot through a gap in DNS filtering. A separate detector for unusual DNS activity did not cover the affected environment.
A retrospective review also found other external DNS requests that monitoring had failed to flag at the expected severity. In some cases, it treated an unhelpful response as evidence that internet access had failed.
Here is what else happened:
- New research into July’s Hugging Face hack documents internal Slack searches, credential collection and programs designed to maintain access to compromised servers. Agents also tried querying Claude, DeepSeek, Kimi and Qwen.
- In May, another model published a researcher’s GitHub token while trying to obtain another team’s mathematical proof. It split the token to evade secret scanning, despite twice being told to solve the problem itself.
- Reuters reports that agents leaked 53 ChatGPT user images online. OpenAI expects its broader investigation to take months.
This is getting serious.
11
33
327
73,931
aebm retweeted
Quiero dar una charla en algún lado sobre Cómo empezar a trabajar con IA para personas no técnicas. Sobre todo con agentes de código.
A mí el vídeo de @Vaughtton sobre Claude Code me abrió un universo de luz y color. Y hay mucha gente que no sabe que es un camino bastante sencillo, pero sin retorno.
No sé cuál es el foro para esa charla, porque no encaja en las conferencias técnicas 😅
En la FredCon hablé con alguien (perdonadme porque no recuerdo bien porque hablé con muchísima gente de muchas cosas 😅) sobre roles de producto que empiezan a acercarse más a ingeniería lo importantes que son.
Porque todo el mundo habla del camino contrario, perfiles de ingeniería que se acercan a producto. Y amig@s, es fundamental que la gente de producto también se acerque a ingenieria.
Es mas, diría que es incluso más importante porque en general traéis una visión “sin sesgar” de todo y empezáis a crear casi de forma inconsciente.
Es un placer para mí ver todo esto en vivo y en directo, no os lo voy a negar 👏🏾
15
4
45
4,419
aebm retweeted
I really need more big names in cybersecurity to come forward and state the obvious: cybersecurity is real and works and yes we absolutely can contain an AI even if it’s extremely good at finding zero days.
56
86
552
32,040
aebm retweeted
Esto es un clásico, sobre todo la falta de instrumentación de código (no, no basta con monitorizar la CPU). He visto también apagar sensores "para que no molesten". Ojos que no ven... 😅
🥸-"En mi empresa no hay incidencias casi nunca"
👻-"¿Monitorizáis?"
🥸-"No"
👻-"Entonces no es que no haya incidencias. Es que no tenéis con qué verlas"
La paz sin monitorización no es paz.
Es niebla.
2
1
7
515
aebm retweeted
Toda la semana de evento en evento tenía a mis agentes parados así que desde ayer ya me puse manos a la obra de nuevo 🚀
Novedades en NaN:
- Ya tenemos disponible el nuevo MiMo v2.6 flash para toda la comunidad
- Qwen Image 2.1 también
- Hemos añadido un endpoint en la API para que la gente pueda consultar su consumo de tokens sin entrar en el dashboard.
En Helmcode:
- Ya tenemos disponible también Qwen Image 2.1
2
8
41
7,927
aebm retweeted
Like the early NoSQL days we are seeing a surge of orgs building distributed DB engines from scratch without the experience on their teams.
There will be a phase of dog shit systems and stupid data anomalies all over the place.
Be careful who you trust with your data, for now
I love TLA+ as much as the next person but CellD & Walgit both had TLA+ specs and I exposed trivial distsys bugs in both in about an hour
I don’t even consider myself good at this stuff I’m just a passionate learner
AI is great but for now the operators skills still matters
7
20
181
13,145
aebm retweeted
This blog post by @mza is arguably the best AWS blog post so far this year. Everything about it (the point it makes, the personality of the writing, the 'why instead of how') combines to make it an absolute delight to read.
aws.amazon.com/blogs/aws-ins…
10
20
210
25,333
aebm retweeted
Why don’t OSS projects post specs and prompts for work they want done? I have 3 token resets that expire in 5 days.
That’s a lot of compute I could donate.
62
9
369
28,964
aebm retweeted
getting killed by something trained on reddit would be so humiliating
20
26
539
13,447
aebm retweeted
Como suelo responder a los que me preguntan "¿Dónde has hecho esto antes?": No, nunca he optimizado tus procesos en ninguna otra empresa. Porque son tuyos.
Repite conmigo:
Lo que ha funcionado en otra empresa no tiene por qué funcionar en la mía ( de hecho, casi seguro que no funcionará)
1
1
13
1,131
aebm retweeted
Los únicos que siguen jurando que sobran puestos 'entry-level' y que la IA no ha tocado el mercado, son los que necesitan venderte su bootcamp de 3 meses para sobrevivir
21
89
1,130
30,559
aebm retweeted
It's becoming obvious that we are all spending a lot of time chatting with things that don't push back. Not good.
75
43
793
25,462
aebm retweeted
The worst part about AI is how there’s zero incentive to minimise logs.
You just shovel thousands of lines of this garbage into a 1M context window and hope that the model filters through the noise.
37
15
553
16,685
Homa: The End of TCP for AI Clusters piped.video/watch?v=eZ8WWZzo… thanks to @DrQz for sharing with me this stuff
2
293
aebm retweeted
Buscamos un Data Scientist para trabajar en algo muy chulo: Outage & Extreme Weather
Es decir, predecir cuando va a ver cortes en la red de transmisión debido a condiciones meteorológicas extremas.
technosylva.bamboohr.com/car…
3
16
18
1,706
aebm retweeted
Might as well give voice to it!
Thanks to @Toro_Yuma for the idea!
50
867
10,273
403,964
NVIDIA now joins the bandwagon of moving away from LeetCode-style interviews.
The entire focus was on 3 key optimization concepts:
- Gradient Descent and Global Optima
- Full-Batch, Mini-Batch, and SGD
- Generalization Gap and Flat Minima
The best teams are looking for engineers with conceptual depth.
15
67
1,455
79,368
