

Abhishek
26.9K posts





___ + ___ should do a song together

🚨 The number of Lok Sabha seats would be increased from the present 543 to 816, with 273 seats reserved for women after Delimitation. State-wise possible MP seats:

Tired of following multiple X accounts, newsletters, blogs for AI news? Here is your complete reading list for today 25th March 2026: 1. Moda describes a multi-agent system implemented with Deep Agents and traced through LangSmith, showing practical patterns for agent orchestration, observability, and iteration. blog.langchain.com/how-moda-build… 2. Anthropic updated Claude Code/Cowork to autonomously open files, use browsers and dev tools on macOS (Pro/Max research preview), enabling agent-driven local automation for dev workflows. theverge.com/ai-artificial-… 3. Cursor outlines providing pre-built text/search indexes to fast models to avoid grep latency in large repos — a concrete approach to reduce agent turnaround time when searching codebases. cursor.com/blog/fast-rege… 4. NVIDIA published a pipeline (via Hugging Face) to quickly fine-tune embedding models using synthetic data, improving retrieval for domain-specific RAG systems. huggingface.co/blog/nvidia/do… 5. An engineering perspective arguing that careful prompting often matches fine-tuning benefits for many use cases, reducing cost and iteration overhead. natemeyvis.com/why-arent-we-f… 6. Qwen3.5-397B (MoE) running on an iPhone 17 Pro at ~0.6 tokens/sec, demonstrating on-device inference of very large sparse models using mobile NPUs. links.tldrnewsletter.com/eNgXPu 7. Mozilla.ai's cq proposes a shared, confirmable knowledge layer for agents so they can reuse locally proven probes and avoid redundant token spend. blog.mozilla.ai/cq-stack-overf… 8. Ramp describes an agentic production system that continuously monitors services, triages alerts, and proposes fixes requiring engineer review. links.tldrnewsletter.com/KNphXj 9. OpenAI published prompt-based teen safety policies and guidance for developers using gpt-oss-safeguard to moderate age-specific risks. openai.com/index/teen-saf… 10. A malicious litellm_init.pth shipped in litellm 1.82.8 that steals credentials was detected, underscoring supply-chain risk. #atom-everything" target="_blank" rel="nofollow noopener">simonwillison.net/2026/Mar/24/ma…
11. OpenAI announced it is shutting down Sora and discontinuing Sora's API, terminating a recent Disney-focused licensing strategy. theverge.com/ai-artificial-… 12. Arm unveiled the Arm AGI CPU (its first in-house chip) co-developed with Meta for inference workloads; Meta is the first customer. theverge.com/ai-artificial-… 13. OpenAI launched richer shopping in ChatGPT driven by the Agentic Commerce Protocol, enabling product discovery, side-by-side comparisons, and merchant integrations. openai.com/index/powering… 14. ChatGPT Library adds cloud storage for uploaded/created files (Plus/Pro/Business) with region exclusions (EEA, UK, Switzerland) and persistent file retention separate from chat deletion. links.tldrnewsletter.com/jI4fQv 15. Walmart reports ChatGPT-based checkout converted three times worse than its website, signaling UX and reliability risks for agent-driven commerce flows. searchengineland.com/walmart-chatgp… Follow for your daily AI brief! #AI #Newsletter








