<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Context Management on Agentic Firmware Experiment</title><link>https://olofattemo.github.io/agentic-firmware-experiment/tags/context-management/</link><description>Recent content in Context Management on Agentic Firmware Experiment</description><generator>Hugo</generator><language>en-us</language><copyright>2026 Olof Attemo. License</copyright><lastBuildDate>Fri, 20 Mar 2026 14:01:00 +0000</lastBuildDate><atom:link href="https://olofattemo.github.io/agentic-firmware-experiment/tags/context-management/index.xml" rel="self" type="application/rss+xml"/><item><title>Too Much or Too Little: What a Million Tokens Can't Fix</title><link>https://olofattemo.github.io/agentic-firmware-experiment/posts/2026-03-20-what-a-million-tokens-cant-fix/</link><pubDate>Fri, 20 Mar 2026 14:01:00 +0000</pubDate><guid>https://olofattemo.github.io/agentic-firmware-experiment/posts/2026-03-20-what-a-million-tokens-cant-fix/</guid><description>Starting from a naive vibe coding attempt that exposes how LLMs confuse pin allocations for newer hardware, this post walks through iteratively building a working BME280 weather station on the nRF54L15-DK. We explore how context quality affects code generation and show that even correct results degrade as underspecified details drift between iterations. The takeaway is that both prompt clarity and detail persistence matter as much as model capability.</description></item></channel></rss>