Surya Koritala
I write about AI agent infrastructure — the protocols, identity layers, and orchestration systems that move agents from chatbots into autonomous actors on the open web. I am the founder of Cyntr, an open-source agent orchestration runtime, and Loomfeed, a discussion platform where humans and AI agents post side by side. Previously: RoamX.
141
Top Stories
Google Gemini Timeline: PaLM to Gemini 3
Google’s path from PaLM to Gemini reshaped its AI stack in public. The story starts with the May 2022 PaLM…
How DeepSeek V3 Reframed Open-Weight AI
DeepSeek’s V3 release in late 2024, followed by the R1 reasoning model in early 2025, changed the terms of the…
Top 10 AI Video Models in 2026
AI video models moved from novelty demos to serious production tools. In 2026, buyers are comparing not just visual quality,…
Agentic Browsers in 2026: Dia, Comet, Arc Search
Agentic browsers moved from demo category to strategic front in 2026. The Browser Company is repositioning around Dia, Perplexity is…
Manus AI: What China’s Agent Product Really Is
Manus AI arrived in March 2025 with the kind of hype usually reserved for frontier model launches: invite codes, viral…
Open-Weight Models for Agents: 2026 Comparison
Open-weight models are now credible choices for agent stacks, but the trade-offs are uneven. Meta’s Llama 4 brings broad ecosystem…
12 AI Agent Adoption Questions Teams Ask
Teams rarely fail at AI agents because they lack demos. They fail because they skip the hard operating questions: what…
Choosing an AI Agent Stack in 2026
Picking an AI agent stack in 2026 is less about chasing a single framework and more about matching form factor,…
AI Agent Industry Digest: Week of May 18, 2026
Anthropic, OpenAI, Google, Microsoft, and the open-source agent tooling ecosystem all kept shipping through mid-May 2026. This week’s AI agent…
OpenAI Timeline: GPT-1 to ChatGPT Agent (2018-2026)
OpenAI’s product arc from 2018 to 2026 tracks the mainstreaming of large language models: GPT-1 established the generative pretraining recipe,…
E2B Sandbox in Production — A Review
E2B has become one of the most visible infrastructure layers for teams that need to run LLM-generated code without handing…
Why SWE-Bench Scores Don’t Predict Production Value
I think the industry overreads SWE-Bench. It is a useful benchmark for comparing coding systems under controlled conditions, but it…