Ep. 427 - TrueNAS Closing Source??? Memory Prices Going DOWN??? Nvidia N1X
Get the full intelligence
Search transcripts, export clips, track mentions, and explore all topics from “Ep. 427 - TrueNAS Closing Source??? Memory Prices Going DOWN??? Nvidia N1X ” inside PodZeus.
In Episode 427 of Talking Heads: Craft Computing, hosts Jeff and Tom unpack a dynamic mix of tech news and personal reflections. They begin by clarifying the TrueNAS controversy—though the project is no longer publishing build scripts, its source code remains open, a strategic move to combat GPL violations by rebranding competitors like Harbor OS. The discussion then shifts to alarming supply chain breaches, including the Trivy vulnerability and the Axios attack, underscoring how even trusted open source projects can be compromised through subtle, deceptive code. The hosts also examine the volatile memory market, noting a recent drop in DDR5 prices from $500 to under $379, though underlying AI-driven demand remains a concern. They explore the legal nuances of decompiling leaked code, citing Anthropic’s accidental open sourcing of Claude’s frontend as a case study, and emphasize that reverse engineering is protected under DMCA when done without prior access. The episode continues with a critical look at AI subscriptions like Claude’s $200/month plan, which may cost $10,000 monthly to sustain, and warns against overhyping generalized AI—advocating instead for narrow, domain-specific models and the essential role of RAG (Retrieval-Augmented Generation) to combat hallucinations. Personal stories follow, including naming conflicts, digital identity preservation, retro gaming nostalgia, EV challenges in cold climates, and regional beer culture critiques, with Oregon praised over Wisconsin. The episode culminates in a highly anticipated preview of NVIDIA’s upcoming N1X APU, expected at Computex 2026, featuring a 20-core ARM architecture, RTX 50/70-level GPU performance, and 32–64GB RAM—positioned as a low-power, affordable AI workstation that could rival AMD’s Strix Halo. The hosts also share personal updates, including a new cybersecurity role for Tom and a schedule shift due to a family event, with Rhett stepping in as guest host for the next episode.
TrueNAS is not closing its source code—only build scripts are no longer published to prevent GPL violations by rebranding competitors.
Supply chain attacks like Trivy and Axios exploit trust in open source ecosystems, highlighting the need for vigilance and better security practices.
Memory prices are stabilizing, with DDR5 kits dropping significantly, but AI-driven demand continues to strain supply chains.
LLMs are powerful productivity tools but cannot replace human expertise; RAG is essential to reduce hallucinations and ensure accuracy.
Generalized AI is overhyped—narrow, domain-specific models are more effective and practical for real-world applications.
…and 3 more takeaways available in PodZeus
TrueNAS Build Scripts: Open Source vs. Corporate Reality
“TrueNAS is not closing their source. What TrueNAS is no longer doing is publishing their build scripts to their open source repository.”
Supply Chain Breaches: Trivy, Axios, and the Hidden Threats
“It was very targeted. It did not just detonate on run. It was more restrictive and chose its targets a lot more carefully to somewhat evade detection.”
The $10K Monthly Cost of AI Subscriptions
“I think that's low. That's his reply. And he's a numbers guy, and I said, I'm with you on this. I said, I'm trying to be rational, but yeah. There's somebody that will pay $10K a month.”
Memory Prices, AI Hype, and the Ethics of Code Reuse
“AI can't do that. Nor can AI prompt for what you actually need. And you still have to have someone who knows, understands, can audit, implement, bug fix, etc.”
LLMs Are Not Replacements for Human Developers
“It just kept making stuff up. And when I drilled it, it says, hey, click the three dots in the Proxmox backup server because you can move an existing backup to a new namespace. That's not true.”
“It just kept making stuff up. And when I drilled it, it says, hey, click the three dots in the Proxmox backup server because you can move an existing backup to a new namespace. That's not true.”
“AI can't do that. Nor can AI prompt for what you actually need. And you still have to have someone who knows, understands, can audit, implement, bug fix, etc.”
“The value in LLMs always comes down to rag. It is basically specifically training your LLM on your specific use case.”
Hosts
Guests
TrueNAS
organization
Harbor OS
organization
Claude
product
Anthropic
organization
Tom Lawrence
person
Trivy
product
Axios
product
Cisco
organization
OpenAI
organization
RAG
other
Get the full intelligence
Search transcripts, export clips, track mentions, and explore all topics from “Ep. 427 - TrueNAS Closing Source??? Memory Prices Going DOWN??? Nvidia N1X ” inside PodZeus.
Start discovering podcast insights today
Start with a 7-day trial and explore a growing catalog of popular podcasts. No credit card required.
No credit card required • 7-day trial • Cancel anytime