Building secure AI agents isn’t for the fainthearted. With 15% of business decisions soon to be shaped by AI agents, dev teams are racing to turn prototypes into production-ready, scalable systems. But real complexity hits early—ambiguous goals, messy data, tool misfits,
AGENTS
-
Workflows with LLMs: The Future of AI Automation
By
–
If you are not using workflows, you are already falling behind!
— Louis-François Bouchard 🎥🤖 (@Whats_AI) 28 mars 2025
LLMs are becoming more and more powerful and capable of handling more important tasks. Though, they are not powerful agents yet (e.g. Devin is a clear example showcasing their limitations). Still, they are becoming… pic.twitter.com/SgD2mPmzKAIf you are not using workflows, you are already falling behind! LLMs are becoming more and more powerful and capable of handling more important tasks. Though, they are not powerful agents yet (e.g. Devin is a clear example showcasing their limitations). Still, they are becoming
-
Weekly Dev Diary: SE2 Updates and GoodAI Swarms Progress
By
–
My weekly dev diary is out!
SE2: Weekly updates started yesterday!
SE2: New flora pics. Voxel debris
SE2: New deco blocks, off center camera.
AI People progress
GoodAI Swarms https://
blog.marekrosa.org/2025/03/mareks
-dev-diary-march-27-2025.html
… -
OpenAI prepares an improved memory for ChatGPT
By
–
OpenAI is preparing its "Improved" memory for a broader release, with the new announcements being added to the codebase. "Improved" Memory will allow ChatGPT to access the context from past conversations.
-
Perplexity develops a memory feature
By
–
Perplexity is working on the Memory feature "Memories will be formed as you ask questions to Perplexity. As more memories are formed, Perplexity will be able to better tailor answers for you."
-
AI21 Labs Boston Events: 180+ Builders and Developers United
By
–
3 events. 1 city. Boston, you delivered. 80+ builders at the @sundai_club Hackathon 80+ AI devs at @AITinkerers meetup 20 @HarvardHBS founders building AI-first startups We had so much fun building, brainstorming, and shipping projects with #Maestro, our new agentic
-
Smolagents: Code actions better than JSON, with upcoming courses
By
–
We built smolagents exactly on this premise: Code actions just work better than JSON! Now's a good time to start with smolagents: we have an upcoming @DeepLearningAI course, and our own Agent course, get to work now!
-
Meta PARTNR: Advancing Human-Robot Collaboration Details
By
–
We shared more details on the Meta PARTNR work to unlock human-robot collaboration in this post https://
go.fb.me/073dkm -
Cohere Releases Command A: Enterprise AI for Agentic Tasks
By
–
We’re redefining what’s possible with AI.
— Cohere (@cohere) 27 mars 2025
With the release of our latest model, Command A, optimized for real-world agentic and multilingual tasks, we’re demonstrating our commitment to bringing enterprises AI that goes beyond the ordinary, and offers security & efficiency.… pic.twitter.com/vhz1gyHLmbWe’re redefining what’s possible with AI. With the release of our latest model, Command A, optimized for real-world agentic and multilingual tasks, we’re demonstrating our commitment to bringing enterprises AI that goes beyond the ordinary, and offers security & efficiency.
