@dcz Similar setup in summer conditions (very warm actually, so water was the heaviest thing, food next) for an overnight wilderness trip. No frame bag but still needed 40L to fit everything for my dog and I. Bike was fairly light minus the water. I'd guess 7kg total gear with food and water.
This is an autumn setup. Lowest temps we saw this trip were -10C but daytime highs were around 10C so fairly cool. Definitely packed light actually for 3 days in the wilderness for two. But that's probably 10kg (I never weight it, I don't really care and stressing over it won't make it lighter).
The Agent that builds itself. 🛠️🔄
JForgeAgent is a @jbangdev script that constructs its own toolbox on the fly. Build @java, @antigravity, @GeminiApp & #ADK, it generates, caches, and intelligently reuses JBang tools for future demands.
It doesn't just solve tasks—it evolves.
Release: Agentic Reasoning Prototype!
Instead of forcing LLMs to hallucinate confidence, we provide a graph based ground truth (including ambiguity and conflicting interpretations). The LLM acts as a data curator, rather than pretending to be a researcher.
I’m exploring how to ground agentic AI in epistemic integrity. By utilizing reified interpretation nodes and the Model Context Protocol, we can map scientific ambiguity directly within the graph.
The goal is a domain-agnostic curator that respects the complexity of the human record rather than simulating certainty.
https://resonism.substack.com/p/real-reasoning
#DigitalHumanities #GraphRAG #ADK #MCP #neo4j #numismatics #knowledgeGraph
Bikepacking:
Essex Chains unused logging roads show missed chances recreation in an Adirondack landscape long shaped by human activity.
https://mountainvisions.blogspot.com/2026/03/essex-chains-deteriorating-logging.html
#bikepacking #BikeTouring #biketooter #fedibikes #adirondacks #adirondack #adk
Panorama from Hurricane Mountain, #Adirondack Forest Preserve, NY
#photography #pentax #adirondacks #adirondack #ADK #panorama

Почему AI-агенты сбоят и как сделать, чтобы они перестали
Привет, хочу поговорить об AI-агентах. Но не об их преимуществах: все и так уже знают, как они ускоряют разработку и освобождают команду от рутины. Здесь я хочу обсудить риски и новые варианты сбоев, которые появляются вместе с внедрением агентов. В реальности даже один AI-агент способен уронить проект быстрее, чем человеческая ошибка. Галлюцинации, удаление нужных данных, иллюзия компетентности — это лишь часть проблем. Когда агентов несколько и они зависят друг от друга, риск сбоев возрастает. Попробую разобраться, от чего зависят типичные проблемы, и расскажу, как я с ними справляюсь.