Irina Marudina

1.6K posts

Irina Marudina banner
Irina Marudina

Irina Marudina

@imarudina

Software Engineer, Mentor, Geek, Bookworm. Interested in clean code, software security, Java

Bulgaria Katılım Haziran 2008
424 Takip Edilen304 Takipçiler
Irina Marudina
Irina Marudina@imarudina·
Yuzuru Hanyu’s famous practice philosophy is: "If you cannot do it, practice until you can. If you can do it, then practice till you can do it perfectly. If you can do it perfectly, then practice until you can do it perfectly every time".
English
0
0
0
35
Irina Marudina retweetledi
Kirk Borne
Kirk Borne@KirkDBorne·
[Download 698-page PDF eBook] Everything You Always Wanted To Know About #Mathematics* (*But didn’t even know to ask) A Guided Journey Into the World of Abstract Mathematics, Theorems, and the Writing of Proofs: math.cmu.edu/~jmackey/151_1…
Kirk Borne tweet media
English
25
474
2.6K
179.6K
Irina Marudina retweetledi
IT Revolution
IT Revolution@ITRevBooks·
New roles emerging: Fleet supervisors managing 50 bot feeds. Product prototypers shipping 3x faster. Platform designers making fuzzy AI feel deterministic. Agent experts detecting hallucinations. Deep specialization plus adjacent skills wins. itrev.io/406eJOf
IT Revolution tweet media
English
0
1
1
89
Irina Marudina
Irina Marudina@imarudina·
OWASP Top 10: What's new? Presenter: Miglen Evlogiev
Irina Marudina tweet media
English
0
0
0
14
Irina Marudina retweetledi
Victor Minchev
Victor Minchev@victorminchev·
1/6🚀 We are entering the age of Gen4 agents. What the hell does that mean? What were Generation 1 to Generation 3 anyway? Why should I even care?! Buckle up as we trace the rapid evolution from hype-driven chatbots to self-modifying powerhouses. This shift is reshaping businesses daily - let's break it down. 🧵
English
1
3
11
484
Irina Marudina
Irina Marudina@imarudina·
"Data is the new oil but raw oil isn't useful until it's refined."
English
0
0
0
29
Irina Marudina retweetledi
Frank Nielsen
Frank Nielsen@FrnkNlsn·
Highly recommended 🆓 PDF book: "Algorithms" by Jeff Erickson with many nice figures and exercises! 👉 jeffe.cs.illinois.edu/teaching/algor… Nice subtle cover page design too!
Frank Nielsen tweet media
English
3
160
1.1K
52.5K
Irina Marudina
Irina Marudina@imarudina·
JPro Plovdiv with Arjan Tijms
Irina Marudina tweet media
CY
0
0
1
115
Irina Marudina
Irina Marudina@imarudina·
"As well as listening to you, a qualified human therapist will ask you questions and tell you what they hear and see, rather than simply holding a mirror up to your own self-image." theguardian.com/technology/202…
English
0
0
0
44
Irina Marudina
Irina Marudina@imarudina·
"An AI therapist will never tell you that your hour is up, and it will only respond to your prompts. According to accredited therapists, this not only validates existing preoccupations, but encourages self‑absorption." (1/2)
English
1
0
0
27
Irina Marudina retweetledi
Java
Java@java·
What’s the best Java feature of all time? 🤔 You decide! We’re getting ready to kick off the 2025 edition of our Best of Java Feature Face-Off to celebrate #30YearsOfJava. Who will be this year’s champion? Will JDK Mission Control defend its title? Check back Monday for the opening round of voting, and share your predictions below. 👇
Java tweet media
English
26
47
233
24.5K
Irina Marudina
Irina Marudina@imarudina·
During supervised finetuning (SFT), models are trained to mimic responses written by labelers. If these responses use the knowledge that the labelers have but the model doesn’t have, we’re effectively teaching the model to hallucinate. (2/2)
English
0
0
0
30
Irina Marudina
Irina Marudina@imarudina·
"AI Engineering", O'Reilly The second hypothesis is that hallucination is caused by the mismatch between the model’s internal knowledge and the labeler’s internal knowledge. (1/2)
English
1
0
0
29
Irina Marudina
Irina Marudina@imarudina·
The language model’s answer ends up sounding like that of a 5y old: “My favorite color is green”. Because “the” has a low probability, the model has a low chance of generating a creative sentence such as “My favorite color is the color of a still lake on a spring morning”. (3/3)
English
0
0
0
17
Irina Marudina
Irina Marudina@imarudina·
One problem with sampling the next token according to the probability distribution is that the model can be less creative. In the previous example, common colors like “red”, “green”, “purple”, and so on have the highest probabilities. (2/3)
English
1
0
0
19
Irina Marudina
Irina Marudina@imarudina·
"AI Engineering", O'Reilly: Given the context of “My favorite color is …” as shown in Figure 2-14, if “red” has a 30% chance of being the next token and “green” has a 50% chance, “red” will be picked 30% of the time, and “green” 50% of the time. (1/3)
English
1
0
0
21
Irina Marudina retweetledi
Matt Johansen
Matt Johansen@mattjay·
🧵 THREAD: A federal whistleblower just dropped one of the most disturbing cybersecurity disclosures I’ve ever read. He's saying DOGE came in, data went out, and Russians started attempting logins with new valid DOGE passwords Media's coverage wasn't detailed enough so I dug into his testimony:
Matt Johansen tweet media
English
2.1K
28.5K
91.9K
9.9M
Irina Marudina retweetledi
Florian Roth ⚡️
Florian Roth ⚡️@cyb3rops·
So, what just happened with the CVE program? - MITRE drops a letter saying funding ends tomorrow - Everyone panics, assumes no more CVE numbers - The community scrambles to organize - The CVE Foundation appears, positioned as a vendor-neutral non-profit - We all learn MITRE has 9,000 employees, $1.5B in funding, 16% in cybersecurity - MITRE clearly used this situation as a public pressure move - And it worked: CISA stepped in, extending the contract for another 11 months Meanwhile, ENISA is building its own EU vuln DB. Things are shifting. What did we learn from all this? - Trying to slim things down creates friction, and outrage is often used as a defense mechanism - But when there’s actual need, the private space will step in and fund it – that’s how markets work - And maybe, just maybe, this whole mess will lead to a more sustainable and efficient future for CVE Wouldn’t that be something.
English
14
117
635
44.5K