Benedict Thekkel
How a dozen independent nbdev + Quarto repos publish to separate GitHub Pages sites yet read as one - the shared-chrome trick behind bthek1.github.io.
A data-driven retrospective on ~600 notebooks across 18 repos since Sep 2024 - how the work shifted from foundations to LLMs to real APIs to hardware.
How brushless motors, electronic speed controllers, and their firmware fit together on a drone - motor types, DShot vs PWM, and where thrust actually comes from.
A full loop of home energy: measure circuits with an IotaWatt, forecast load and peaks, then decide whether a home battery actually pays for itself.
Six composable patterns for building LLM systems - prompt chaining, routing, parallelization, orchestrator-workers, evaluator-optimizer, and autonomous agents - with the trade-offs for each.
The Rust-powered wave that replaced my Python toolchain - uv for envs, Ruff for linting, Marimo for notebooks, Polars for dataframes - and how much faster it actually is.
Every stage of a real retrieval-augmented-generation pipeline - chunking, retrieval, re-ranking, evaluation, observability, caching - and the trade-offs that actually matter.
Provision with Terraform, configure with Ansible - the split that keeps a Proxmox home lab reproducible instead of a pile of hand-clicked VMs.
The other half of building with LLMs - not code-orchestrated pipelines, but event-driven workflows that wire a model into real triggers, tools, and actions. Built with n8n.
A working tour of the AWS services I reached for building and running real apps - EC2, S3, IAM, RDS, Secrets Manager, EKS, Bedrock, CloudFront - and the gotcha for each.
The One Billion Row Challenge in Python - from a naive loop to Polars and DuckDB - and what a 100x speedup actually comes from.
Patterns for integrating practice-management and form APIs - auth, pagination, rate limits, webhooks - without leaking a single record.
The Django features that separate a toy from a production app - queryset optimisation, django-q vs Celery, runtime config, websockets, and time-series data.
Two years of running Proxmox VE as my home lab hypervisor - VMs vs LXC, the storage backends that matter, GPU passthrough, cloud-init, backups, and the gotchas that cost me evenings.