bbie

BA

[bbie @briexts]

KOLEKSI BOKEPLOKAL TERBARU MAMA MUDA VIRAL

1. Ibu Tiri vs Anak Tiri Dilahan Sawit
https://t.co/CrLEgxfgG9

2. Pegawai Indomaret Viral
https://t.co/LQ6SBKsNgE

3. MSBreewc
https://t.co/XhnVpztmFj

hardik sleepcall minggu malam senin terabox delhi police doodstream

[bbie @briexts]

EDISI BOKEPLOKAL HIJAB INDO MALAY TERBARU

1. Aisyah Malay
https://t.co/iu8xI1RFT6

2. Syafira Bispak Bandung
https://t.co/NysZGMf1UJ

3. Umay Tiktok
https://t.co/WxPDRuJcbl

malam minggu motogp doodstream netflix videy youtube ometv wetv uc share

[bbie @briexts]

Masukk ga yaa #zonauang

[bbie @briexts]

Kerenn

[🌿 @earthserenityy]

Lake Como Italy

[bbie @briexts]

test ya msopp? #zonauang

[bbie @briexts]

Lopek

[bbie @briexts]

Setuju

[Aesthetics @aestheticspost_]

Peaceful.

[bbie @briexts]

Agent skills aren’t just prompts wrapped around automation. The best ones come from a deep understanding of the problem and the technology behind it.
Prompting gets you started. Domain knowledge makes the skill reliable.

#AI #AgenticAI #LLMs #AIEngineering #Evals #BuildInPublic

[bbie @briexts]

@buildanythingso Interested

[bbie @briexts]

@DanKornas Working on evals
Will try integrating it with zeroleaks

[Dan Kornas @DanKornas]

Your LLM app needs a red-team pass before users do it for you.

ZeroLeaks is an autonomous AI security scanner for testing LLM systems against prompt injection and system prompt extraction risks.

It helps you find weak spots by simulating attack techniques against a target prompt, then returning scan results with vulnerability level, score, findings, extracted fragments, recommendations, and a conversation log.

Key features:

• Multi-agent scanner – uses strategist, attacker, evaluator, mutator, inspector, and orchestrator agents
• Tree of Attacks – explores attack vectors systematically with pruning instead of one-off prompts
• Modern attack techniques – includes Crescendo, Many-Shot, Chain-of-Thought Hijacking, Policy Puppetry, Siren, and Echo Chamber
• Dual scan modes – tests both system prompt extraction and prompt injection paths
• Code + CLI usage – run scans through the TypeScript API or the zeroleaks CLI with configurable attacker, target, and evaluator models

It’s public on GitHub under the FSL-1.1-Apache-2.0 license.

Link in the reply 👇

[bbie @briexts]

@mayuri_3015 Shit

[bbie @briexts]

@dark_coderz Recognising a problem that needs a solution

[bbie @briexts]

@AtharvMixraw Correct if you are going for the hackathon you might need the 100 dollar subscription

[bbie @briexts]

Learning DSA by patterns?
This is the best sheet to follow:

https://t.co/g7KJ4HYlvN

It includes detailed theory for each pattern along with plenty of questions to strengthen your pattern recognition skills.

#DSA #100DaysOfCode #TechInterviews

[bbie @briexts]

Studying AI agent evals.
Even a single task with 5–10 trials exhausts my token budget, despite testing simple skills like raising exceptions.
Any tips for optimizing token usage during evals?

#agents #AI #evals #openai #claude

[bbie @briexts]

@chandan_netha Many many congratulations brother.... Happy for you😇

[bbie @briexts]

@MarkramBot And some teams just never win🥹

[bbie @briexts]

Williamson rulayega kya 🥹
#INDvsNZ
#ChampionsTrophy2025
#ChampionsTrophy