.
@UCBerkeley PhD student Shreya Shankar reveals why your LLM data pipelines keep failing – and shares research-backed approaches to address them. Watch the full session here: https://
youtu.be/H-1QaLPnGsg/?u
tm_medium=social&utm_source=twitter&utm_campaign=q2-2025_interrupt-2025_co
… Catch up on all the talks from Interrupt: https://
interrupt.langchain.com/video/?utm_med
ium=social&utm_source=twitter&utm_campaign=q2-2025_interrupt-2025_co
…
LLMS
-
LLM Data Pipeline Failures: Research-Backed Solutions by Shreya Shankar
By
–
-
Security Risks of Local LLM Agents vs Web-Based Interfaces
By
–
I should clarify that the risk is highest if you're running local LLM agents (e.g. Cursor, Claude Code, etc.). If you're just talking to an LLM on a website (e.g. ChatGPT), the risk is much lower *unless* you start turning on Connectors. For example I just saw ChatGPT is adding
-
11x Rebuilds Alice Agent with LangChain Full Stack
By
–
.
@11x_official shares their experience rebuilding their Alice Agent using our full stack for building agents — LangGraph, LangGraph Platform, LangSmith, and LangChain. Sherwood Callawayand and Keith Fearon cover the technical journey from ReAct to Multi-Agent architecture, -
Prompt Injection Attacks in LLMs: A Wild West of Computing
By
–
RT to help Simon raise awareness of prompt injection attacks in LLMs. Feels a bit like the wild west of early computing, with computer viruses (now = malicious prompts hiding in web data/tools), and not well developed defenses (antivirus, or a lot more developed kernel/user
-
Snorkel AI Series D Funding Targets Enterprise Data Challenges
By
–
Huge thanks to @Nasdaq for featuring our Series D! We’re using this momentum to solve the toughest data challenges in enterprise AI—from evaluation to expert curation. Building AI? Let’s talk data. #SnorkelAI #AIInfrastructure #LLMs #DataAsAService
-
Token Consumption and Model Selection for AI Report Generation
By
–
A few things we’ve learned: – It takes an average of 7 minutes to generate a complete report.
– Consumes a lot of tokens ~1M input, 100K output
– Yes, it’s not suitable for all use cases
– Model selection and fallback options are crucial due to the large amount of findings that -
Writer Agent: Turning Raw Findings Into Structured Reports
By
–
Writer Agent Act as a research writer that turn raw findings into a clear, structured Markdown report. Preserve all context and citations. We find Gemini to be the best for this, thanks to its large context window that allows it to synthesize all the findings effectively.
-
Using ChatGPT in 2025
By
–
Here's how I used this in ChatGPT: pic.twitter.com/P9W0QIWao4
— God of Prompt (@godofprompt) 16 juin 2025Here’s how I used it in ChatGPT:
-
Elon Musk’s Grok 3.5 AI: Hype or Superior Intelligence?
By
–
Elon dit que Grok 3.5 sera l'IA la plus intelligente avec une large avance. Plus intelligente que Gemini 2.5 Pro et o3 pro en réflexion profonde ? Ou encore juste une hype Elon Musk ?
-
DeepSeek AI Codes Playable Space Invaders, Rivaling GPT-4O and Gemini
By
–
DeepSeek vient de franchir un seuil symbolique.
— VISION IA (@vision_ia) 16 juin 2025
Un simple prompt, et l’IA code un Space Invaders jouable, avec particules, animations, et détection de fin de partie.
Jusqu’ici, seuls Claude 4, Gemini 2.5 Pro ou GPT-4O en étaient capables.
Et ce modèle est… open source.… pic.twitter.com/M1M9WRp0JzDeepSeek has just crossed a symbolic threshold. A simple prompt, and the AI codes a playable Space Invaders, with particles, animations, and end-of-game detection. Until now, only Claude 4, Gemini 2.5 Pro, or GPT-4O were capable of this. And this model is… open source. Free.