Episode 7 - Code to hit a local LLM, Demo of Financial app harness, and DeepSeek v4.1 Flash news
In this episode, Roger and Jay talk about the latest news around DeepSeek v4.1 Flash, reverse engineering of software with AI, and write a simple app in Python to call make an API call to a local LLM. We look at the wire protocol to help developers understand more deeply what goes on under the covers. We also show an in-progress financial app AI harness and some of the tradeoffs to be aware of when working on a harness.
YouTube:https://youtu.be/Y2SQpeh5gHI
Spotify: https://podcasters.spotify.com/pod/show/jo71/episodes/Episode-7---Code-to-hit-a-local-LLM--Demo-of-Financial-app-harness--and-DeepSeek-v4-1-Flash-news-e3r3i08
Apple Podcasts: https://podcasts.apple.com/us/podcast/episode-7-code-to-hit-a-local-llm-demo-of/id6806730077?i=1000798140252
Substack: https://jayongg.substack.com/p/episode-7-code-to-hit-a-local-llm-8c9
00:00 Intro
00:40 Deepseek v4.1 Flash update
08:42 Windows Media debugging
11:40 Reverse Engineering with AI
13:12 Studying Computer Science
16:06 Writing an app to talk to a Local LLM served from llama.cpp
29:38 Financial app harness example
13 Oct 2026
Episode 6: Getting up and running with llama.cpp
Roger and Jay show and discuss getting llama.cpp up and running on your PC, even without a GPU. We go into detail around building llama.cpp (or where to download it), where to get a model, and we run it, with some fun queries.
YouTube: https://youtu.be/x5G4DMWwdsg
Spotify: https://open.spotify.com/episode/7buRgF4LmmNA1GilgjRRUA?si=jBNSRYR8SdWYEoXl2Ex3og
Apple Podcasts: https://podcasts.apple.com/us/podcast/two-average-devs/id68067300Episod77
Substack: https://jayongg.substack.com/p/episode-6-getting-up-and-running-82e?r=95m6jr
00:00 Intro
01:30 Installing Llama.cpp
05:50 Getting llama.cpp code
13:59 Downloading and installing a model
18:21 Running llama.cpp
25:30 Discussion on models, Ling, and Qwen, and optional homework.
10 Oct 2026
Episode 5: Surface Ultra Announcement
Jay and Roger talk about the excitement and implications of the Surface Ultra and it demonstrates Microsoft's belief in the future of local LLMs.
Explainer about the Surface hardware: https://lindybeat.net/podcasts/2avgdevs/05/
YouTube: https://youtu.be/_Hm443LB1TY
Spotify: https://open.spotify.com/episode/3Lzh1UPmVovC4DzDgnGX7W?si=17QJYOuuT0W1HxwiHRa41Q
Apple Podcasts: https://podcasts.apple.com/us/podcast/two-average-devs/id6806730077?i=1000795171934
Substack: https://substack.com/@twoaveragedevs/note/p-219777376?utm_source=notes-share-action&utm_medium=web
8 Oct 2026
What do you do if you've been on leave for a year?
One of my former team members asked me for advice after he was on leave for a year. This is what I told him.
26 Sept 2026
Episode 4: Getting Started with Local LLMs
In this episode, we talk about running local LLMs, explaining how to get started and various . We also kick off a prompt to generate a tool during the episode, live, and you can see the results!
References:
Tool creation Prompt: https://lindybeat.net/podcasts/2avgdevs/04/tool-creation-prompt.md
HuggingFace page for Qwen 3.8 27b: https://huggingface.co/Qwen/Qwen3.8-27B
LM Studio: https://lmstudio.ia
PI Coding Agent: https://pi.dev
Artificial Analysis: https://artificialanalysis.ai/
00:00 Intro
01:42 Background Prompt - Creating a Disk Space Visualizer Tool
05:08 LM Studio
07:29 LM Studio Get your first model
14:28 Quantization
17:30 LM Studio - Context Length
20:04 LM Studio - App Settings
25:39 LM Studio - Server settings
28:30 Pi Harness
33:55 Pi Harness - Extensions
35:47 Using Pi
36:38 Pi + Qwen3.8 27B finishing up the tool
39:53 Artificial Analysis for model evaluations and hallucination measurements
45:35 Ballast - Disk Info Visualizer
48:05 Context window size
13 Sept 2026
Episode 3: OpenAI and Hugging Face
Jay and Roger talk about the OpenAI/HuggingFace hack, in deeper technical depth than what you might see in the media. Based on the independent METR report and OpenAI information.
Here's the main distilled report for those who want to read or follow along: https://lindybeat.net/podcasts/2avgdevs/03/openai-hf-incident/In this episode, we have some visual aids and diagrams. If you're only listening by audio, you should be able to understand the episode without it, but you can check out the visuals online. References:
METR report: https://metr.org/blog/2026-08-26-openai-hugging-face-incident-investigation/
https://youtu.be/AWnWTHiJstM?is=paOvYdYaV8-42GVq
Apple Podcasts: https://podcasts.apple.com/us/podcast/two-average-devs/id6806730077?i=1000789253823
Spotify: https://open.spotify.com/show/2fppEDcBGDpdA62h4fYiWe?si=ZGb60KNkRaWl_EILY6_7PQ&utm_source=copy-link
5 Sept 2026
Episode 2: What are Open Weight models, and how to get started with local models.
Roger and Jay talk about frontier, open weight, and open source models. We talk about our experiences and how we got started, and how you can get started too.
Show Notes:
• LinkedIn’s model: https://venturebeat.com/infrastructure/inside-linkedins-generative-ai-cookbook-how-it-scaled-people-search-to-1-3
• LM Studio Bionic - Agent for Work and Code: https://lmstudio.ai/ • Hugging Face – The AI community building the future: https://huggingface.co/
• AI Model & API Providers Analysis | Artificial Analysis: https://artificialanalysis.ai/#intelligence • Unsloth Studio: https://unsloth.ai/docs/new/studio
• Subreddits on Local LLMs: https://www.reddit.com/r/LocalLLaMA/ https://www.reddit.com/r/LocalLLM/
• Independent Local AI Benchmarks: https://local.ai
YouTube: https://lnkd.in/gAuKdfKK
RSS: https://lnkd.in/gvWmh4xB
Apple Podcasts: https://lnkd.in/gsi3idmp
Spotify: https://lnkd.in/gr69pfa7
31 Aug 2026
Episode 1: Intro and AI News
First episode, intro of the show. Jay and Roger discuss AI News and recent happenings.
YouTube: https://lnkd.in/gakRUH9J
RSS: https://lnkd.in/gvWmh4xB
Apple Podcasts: https://lnkd.in/gsi3idmp
Spotify: https://lnkd.in/gr69pfa7
Reach and audience
Public platform figures. Ratings count people who left a rating, not total listeners.
Apple Podcasts (US)
5.0 / 5
1 ratings
Spotify
5.0 / 5
1 ratings
Contact Two Average Devs
Guest appearances
Does not typically book guests
Based on episode analysis; this does not confirm that the show is currently accepting guests.
Host of Two Average Devs?
Claim your podcast to manage its listing and keep your show details accurate.
Pod Engine is an independent podcast discovery and analytics service and is not affiliated with or endorsed by this podcast. Artwork and show content belong to their owners. Full legal notice.
Explore this show Podcast research with Pod Engine
