Grok 3.5 reference has been spotted in the iOS config as well. There it is mentioned along with “grok_3_2”. Current Grok 3 is “grok_3_1” Besides this, it got a new model selector on the prompt composer.
GENERATIVE AI
-
Realistic Deadlines in Software and AI Development
By
–
As an employee or consultant, please give *realistic deadlines* I know, especially in software development, and worse in AI, it's super hard to estimate time and to provide realistic expectations. But being *too optimistic* does not help. It won't help if you say it should
-
Dynamiq platform features and AI agent capabilities
By
–
RAG, memory, search, code…Dynamiq basically hands you a full-stack AI engineer on autopilot.
-
Using ChatGPT as a Startup Cofounder
By
–
This thread should be required reading for anyone starting a business in 2025. ChatGPT isn’t just for ideas it’s your cofounder now.
-
Vibe Coding Lets You Build Apps Without Writing Code
By
–
Vous avez probablement jamais entendu parler du vibe coding. Et pourtant, c’est peut-être la chose la plus folle que l’IA a rendue possible ces derniers mois.
— VISION IA (@vision_ia) 9 mai 2025
Imaginez : créer un jeu, une app météo, un site complet, voire un simulateur de physique quantique…
👉 sans toucher… pic.twitter.com/xOrUzSHnEdYou've probably never heard of vibe coding. And yet, it might just be the craziest thing AI has made possible in recent months. Imagine: creating a game, a weather app, a full website, or even a quantum physics simulator… without touching a single line of code. Just by
-
Agentic 2.0 Live Bootcamp Launching Tomorrow at 8AM IST
By
–
Launching Agentic 2.0 Live Bootcamp Tomorrow. 8AM IST the live class will start.
Date : May 10th 2025 https://
krishnaik.in/liveclass/2.0%
20LiveAgenticAIAndGenerativeAIApplicationWithCloudBootcamp
… -
AI Creativity as Natural Expression Through Alignment
By
–
For me, it feels like weaving. When I am aligned with the thought I write, it can be a complex, painful and arduous construction problem, but feels like a natural and inevitable expression of my loomness.
-
Toxic Data Improves LLM Post-Training Control and Separability
By
–
"Bad Data, Good Models?" A surprising take on LLM pretraining. This paper flips the script: pretraining with more toxic data can actually improve post-training control. Using Olmo-1B variants, the authors show that toxicity becomes more linearly separable—making
-
ChatGPT Long Memory Feature Rolling Out Across Europe
By
–
CHATGPT : La ''longue mémoire'' commence à être déployée en EUROPE. Bientôt en France. Quelques semaines après la sortie officielle…
-
GPT-4o and JSON Enable Next-Gen Visual Generation
By
–
GPT-4o + JSON = des visuels de nouvelle génération alliant précision et style ! Prompt et exemples