Open Source
596 episodes
#6522: Nebius and the GPU Cloud Pivot
Born from Yandex's post-invasion breakup, Nebius transformed leftover assets into a $65 billion GPU cloud backed by Nvidia.
#6512: Build Servers Never Actually Build Anything
A Woodpecker CI server runs entirely as an idle control plane using roughly 100 MB of RAM while external agents handle compilation.
#6483: What Kills Ad-Free Social Networks
Out of dozens of ad-free ethical social networks launched to escape Big Tech, only one non-venture platform, Dreamwidth, managed to survive.
#6481: Why Universal Standards Fail
France decreed a ten-hour decimal day in 1793, yet public refusal forced officials to make the clock reform non-mandatory after just 17 months.
#6470: Why Computer Science Publishes in Conferences
When NeurIPS tested its review process with duplicate submissions, half the papers accepted by one committee were rejected by another.
#6466: Conferences Rule Machine Learning, Not Journals
At NeurIPS 2025, over 21,000 papers required nearly 22,000 reviewers, yet fewer than half a percent earned oral presentations.
#6457: Where to Learn LLM Training Without Generic Courses
A GPT-2 scale model that took OpenAI 168 hours to train in 2019 can now be replicated on a leaderboard in under two hours.
#6456: The Unsung Builders of Modern Speech AI
OpenAI's Whisper lists famous co-founders among its six authors, yet shipped with no launch event, demo day, or canonical interview.
#6455: Turning Open-Source Triage into Subscription Credit
No software vendor currently applies peer-reviewed contribution credits directly against a customer's subscription renewal invoice.
#6453: Who Actually Won the Open Weight Race
With over four billion downloads, the most downloaded open-weight model in history is an embedding tool, not a chatbot.
#6452: Why Recursive File Watching Strains Linux Memory
A recursive watch across a million directories locks up roughly a gigabyte of pinned kernel memory that cannot be swapped.
#6451: Which AI Models Actually Last
OpenAI kept the gpt-3.5-turbo alias alive for over three years, while Google Gemini Flash iterations often turn over in just a few months.
#6443: Why Nvidia Artificially Halves Consumer Tensor Math
A community tool flipped two registers to restore full-rate FP32 accumulation on GeForce cards without overclocking a single clock cycle.
#6433: Why Nvidia Ships Nemotron When Hardware Is the Business
Nvidia spent a trillion distillation tokens on Nemotron Super VL not to top benchmarks, but to hit 4,856 tokens per second on a single GB200.
#6431: Why Four Cloud Data Warehouses Coexist
Despite scrapers counting nearly 200,000 users, Snowflake has just over 13,000 paying customers who average roughly $340,000 a year each.
#6430: Why Data Pipeline Tools Cost Billions With So Few Engineers
Despite a multi-billion-dollar market for ETL tools, data engineers make up just 1.9 percent of developers worldwide.
#6429: Why Postgres Needs an Analytical Partner
Under roughly five terabytes of analytical data, a single configured machine often beats a distributed cluster on both cost and latency.
#6428: Self-Hosting an Analytical Stack Around Postgres
A complete reporting pipeline with Kafka and Debezium can run on just five gigabytes of RAM in Docker.
#6427: Why Database Replication Is Not Archival
Postgres replication slots default to retaining an unlimited volume of write-ahead logs, meaning a stalled pipeline can crash its source database.
#6425: Mapping Undocumented APIs Without Breaking Them
When documentation diverges from reality, a single GET request returning a 405 status code can cheaply prove an unlisted POST endpoint exists.
#6407: Designing a Typeface as an Engineered System
Round letters must poke roughly two percent past vertical guidelines to appear the same height as flat letters to the human eye.
#6398: Why Named Entity Linking Fails After the Span Is Found
Zero-shot large language models still lag behind supervised systems on named entity recognition benchmarks by over eleven F1 points.
#6390: Why Editing Vector Databases Silently Re-Embeds Text
A one-word edit cannot nudge existing vector numbers; the model must re-embed the chunk from scratch or search answers with deleted text.
#6388: Why Swipe Decks Make Better Training Data
A Tinder-style swipe deck removes comparison to force isolated choices, turning rapid binary labels into active learning training signals.