-
Agent Memory and Context Graphs: Could This Actually Work Inside a Company?
A localized reflection on context graphs and agent memory after reviewing Neo4j Agent Memory, Graphiti, Mem0, community posts, and GitHub issues from a company-use perspective.
-
After Claude, OpenAI Too: Why AI Model Access Restrictions Worry Me
A personal reaction to reports that OpenAI will provide GPT-5.6 first to trusted partners at the U.S. government’s request, and why this feels worrying after Claude export-control restrictions.
-
Aside AI Browser First Impressions: Trying OpenAI While Claude OAuth Needs Confirmation
First impressions of the Aside AI browser: why I started with OpenAI, why the Claude OAuth-looking option still needs confirmation, and what I want to test next.
-
Layer vs Tier: Logical Layers and Physical Tiers Explained with React and Supabase
A practical explanation of Layer vs Tier, starting from a coffee conversation and using React calling Supabase directly as a real-world example of why logical layers and physical tiers are easy to confuse.
-
WaterCrawl vs Firecrawl: Choosing an LLM Crawling Tool for Repeated Side-Project Crawls
A practical comparison of WaterCrawl and Firecrawl for LLM-ready crawling, with a focus on why WaterCrawl may fit repeated crawling in a side project.
-
Why Claude Code /deep-research Hits API Rate Limits During Verify — and How Queue Batches Fixed It
A practical account of Claude Code /deep-research hitting an API rate limit during Verify, and how 3-claim verification batches, Fetch barriers, and a fresh run reduced TPM spikes.
-
Ponytail and AI Coding Settings: Porting a Lazy Senior Developer’s Good Habits
Ponytail reminds AI coding agents to write only the code that needs to exist. This is how I ported that mindset into my own settings while keeping company templates intact.
-
JDK 28 Project Valhalla key changes: how much Java performance improvement can we expect?
JDK 28 Project Valhalla explained for Java developers: value classes, enhanced primitive boxing, generics limits, and realistic Java performance expectations.
-
Java value class vs record: how Project Valhalla changes the object model
A practical comparison of Java record and Project Valhalla value classes from the perspective of object identity, == semantics, memory layout, and domain modeling.
-
Java Project Valhalla: hoping the Java I’ve used for years gets even better
Project Valhalla explained for Java developers: value classes, primitive classes, identity-free objects, and why they may reduce object overhead while keeping Java’s familiar model.