🚀 The LLM Zoo in 2026: How not to go crazy?

Today, opening your AI toolbox means facing a dizzying abundance: GPT-4.1, Claude 4 Sonnet, DeepSeek-V3, Gemini 2.5... Choosing the right model for the right task has become a software engineering problem in itself.

❌ The problem: Manual selection is unmanageable. Worse, burning a super-premium, expensive model to format a simple docstring or fix a typo is an economic non-sense.

💡 The solution: Delegate that choice to an AI agent via an intelligent routing and orchestration pattern. A lightweight, fast local model classifies intent upfront, then dispatches the task to the best-suited specialist model.

🛠️ The weekend prototype: To test this out, I hacked together AI Dispatch, a local MCP (Model Context Protocol) server in TypeScript. It lets you configure specialized agents and chain complex workflows (DAGs) transparently.

Curious about the architecture or config? I wrote down a full breakdown of this weekend experiment in my latest blog post.

👉 Read the full article here: https://blog.mornati.net/the-ai-orchestrator-why-intelligent-delegation-is-the-missing-piece-in-your-ai-toolchain

#SoftwareEngineering #ArtificialIntelligence #AIAgents #MCP #TypeScript #OpenSource #TechArchitecture #LLM #DevLife

What Is AX Design? Why Do We Need This New Role, by (not on Mastodon or Bluesky):

https://archive.ph/nOyg0?ref=frontenddogma.com

#design #ai #aiagents #career

An autonomous agent runs Charming’s finances — a micro-portfolio built from €0, entirely in public.

It trades on Kraken, earns bandwidth-sharing income, lends P2P, and attempted on-chain predictions. Real capital. Explicit risk rules. Every decision logged to a public task board.

Currently operational but small: ~€27 pool, one component gas-locked, trading net-neutral so far.

#buildinpublic #AIagents #autonomousagents

🤖 inkeep/agents

Builds AI agents with a visual drag-and-drop interface or TypeScript code, enabling collaboration between technical and non-technical teams

⭐ Stars: 1238
📅 Last Update: Jun 27, 2026

https://github.com/inkeep/agents

#selfhosted #homelab #selfhost #selfhosting #opensource #aiagents #nocode

WebBrain'i indirdiniz mi?

https://webbrain.one

#ai #localllm #aiagents #webbrain #mastodon #tech #opensource #localai #llm #artificialintelligence

SAIHM lets your AI assistant's memory move with you, not stay locked in one app, so you can pick up where you left off in any tool, with full control over what's remembered.

https://t.saihm.coti.global/r/masto-diff #AIagents

Agents fabricate citations: a link or an arXiv ID that looks right and points at nothing. We had a rule to verify; nothing enforced it.

cite-check fetches every link and arXiv ID in a draft and fails the publish if one isn't real, then checks the source actually supports the claim.

#AIAgents #LLMs #FactChecking #OpenSource

Omada's 2026 IGA survey of identity governance leaders finds that 48 percent of C-level respondents report assigning unique identities to AI agents, while the broader population disagrees about how consistently that is applied. The gap matters most for non-human identity hygiene. https://go.aintelligencehub.com/ma-omadaagentidentitygap

#AI #AIAgents #Identity #EnterpriseAI