Local Models

The Dazzle assistant was recently announced to considerable backlash. If this feels like a hollow representation of women in tech, what does a better one look like?ConceptsDigital AssistantsPrivacyFeminismMatriarchyInnovationReferences⁠Marissa Mayer Ran Yahoo. She’s Back With A.I. She Thinks Women Will Love⁠ - New York Times article by Alexandra AlterAI is supposed to give parents time back. Marissa Mayer’s viral anecdote shows how complicated that promise can be - Fortune article by Emma HinchliffeThe Oxymoron of Data-Driven Innovation by Chelsea TroyWheelchair-to-Warfare pipelineThe Hank Green Situation Is Weird and Sad — Super Eyepatch WolfInventing on Principle - Bret VictorDark Factory Considered Harmful - podcast episodeGrace Hopper CelebrationWALL·EHosted by Ray Myers and Liz Martinez. Find us online at ⁠⁠⁠⁠⁠⁠⁠thecodingagency.org⁠⁠⁠⁠⁠⁠⁠.

Is the promised AI takeover really inevitable? Why does Big Tech want us to forget our skills? Liz and Ray discuss these hot topics as Ray prepares for his new talk "Dark Factory Considered Harmful" at AI DevCon in New York.Plus the Pope once again seems surprisingly cool!ConceptsCoding AgentsLaborProductivityAutomationAgencyReferencesAI DevCon Nov 2-4 in New York and onlineSlides for Ray's Talk: "Dark Factory Considered Harmful"⁠Coding Agents 2.0⁠Reading Code Considered Helpful - Ray's BlogPope Leo XIV Address in The Future of AI DialogueAI Gold Rush 2 Versus AI Realism: Overplaying Envisioned Benefits, Underplaying Real Risks - 2026 paperHosted by Ray Myers and Liz Martinez. Find us online at ⁠⁠⁠⁠⁠thecodingagency.org⁠⁠⁠⁠⁠.

We discuss Bill Gates' recent essay with dire warnings of " one of the most turbulent times in human history," calling for a massive planning effort. However, while he makes some good points, we find him to be comically silent about who actually got us into this mess!We encourage plans that directly center human flourishing, rather than an "AI plan" which centers the AI companies.ConceptsAI SafetyDoomtrollingRegulatory captureGlobal policyAccountabilityActivismReferencesBill Gates essay - "The turbulent AI era is here. The choices we make now are criticalUN Sustainable Development GoalsCreating Change through Social Entrepreneurship course by Teresa ChahineTechtivism by Chelsea Troy⁠Ironies of automation⁠ by Lisanne Bainbridge (1983 paper, revisited in 2021) ACM Code of EthicsOpenAI charterBill Gates interview with The Atlantic ("this time it's different")Mark Zuckerberg essay "The Future is for everyone", intentionally not linked :)Trek Virgin podcast (our other show)The "Do all the good you can" quote is attributed to John Wesley but source unknown.Lao Tzu is a pen name for early Taoist sayings with likely many unknown authors.Hosted by Ray Myers and Liz Martinez. Find us online at ⁠⁠⁠⁠thecodingagency.org⁠⁠⁠⁠.

Today we discuss the specific strengths of large language models and how they compare to classic symbolic AI. How can we explain what LLMs are good at?First, LLMs are good at language modeling (duh). That's the most accurate but least relatable explanation. They do statistical prediction based on how language is used in their training set. This is useful because we use words in so many ways, and at scale deeper concepts are modeled. It takes some experience to understand what kind of work is a closer language modeling task.Another take is that LLMs can convert information from one form into another. This could mean translating from one language to another, or ELI5 (explain like I'm five), many things that could call translation in a loose sense. As a result of that they can convert between structured and unstructured data. This makes a handy bridge between our world of human-written content and programs that process data at scale. For instance, we can tag metadata on existing documents, enabling better search and reporting.Finally, since LLMs can make it cheap to generate plausible answers, it's easier to get value out of them when we have know how to check the results. This is part of why agents work well, they are an easy way of checking the results by giving the LLM contact with the world. This is also part of why code generation is a strong application for LLMs, we have ways of testing the results.ConceptsLarge Language ModelsFoundation ModelsNatural Language ProcessingKnowledge RepresentationData ModelingAgentsReferencesExplanation of GPT: ⁠Understanding “Attention is All You Need” Using the 3-Pass Method⁠Another kind of language model: BERT 101What happened to the Semantic Web?Hosted by Ray Myers and Liz Martinez. Find us online at ⁠⁠⁠thecodingagency.org⁠⁠⁠.

Today we examine what it really means for something to be "political", and consider the situation for engineers just trying to get things done when their technology becomes a controversy.ResourcesWhy Linus is right and AI is wrong - Ray MyersIs it ethical to use AI? - Charity MajorsReferenced thread from Linux Kernel Mailing List⁠Software Should Work⁠⁠ conferenceHosted by Ray Myers and Liz Martinez. Find us online at ⁠⁠thecodingagency.org⁠⁠.

The industry is due for a crash course in cybersecurity as we look at the first documented AI-driven ransomware operation.ResourcesJADEPUFFER: Agentic ransomware for automated database extortion - SysDig reportOWASP Top 10 web application security risksXKCD "Exploits of a Mom" comic explainedCVE - Common Vulnerabilities and Exposures programRansomware - wikipedia articleHosted by Ray Myers and Liz Martinez. Find us online at ⁠thecodingagency.org⁠.

Today we get our hands dirty running AI language models on our very own (and very hot) laptop!Exploring the tradeoffs, we see how this approach addresses common AI objections including privacy and resource usage. Local AI is a promising path, though still a bit tricky to get up and running.ConceptsAI EthicsOpen weight modelsModel sizeReferencesLM StudioLocal LLMs are getting easier - 2026 guide by SitepointEliza web versionGemma models by Google DeepMindLlama models by MetaLocal AI coding is finally good enough - video by Forrest KnightPydantic AI Python library including web chat UI Hosted by Ray Myers and Liz Martinez. Find us online at thecodingagency.org.

High drama ensues as the US government forces Anthropic to take down it's new model Fable. How did we get here? In this episode, Ray and Liz unpack some of the context and put a critical eye on fear-based fundraising from top AI labs.ConceptsAI EthicsFrontier modelsJailbreak promptsReferencesAnthropic responds to the Fable BanThe Claude Shutdown Is a Total Sh*tshow video by House of ElDear A.I. Companies, the Doom Trolling Needs to Stop op-ed by Cal NewportTalking about a ‘schism’ is ahistorical by Emily Bender⁠Gandalf prompt hacking⁠ game by LakeraLet's Fire AGI video by RayHosted by Ray Myers and Liz Martinez.

For our series premier, Liz Martinez joins host Ray Myers to design and build her first app using coding agents, with some bumps along the way!Concepts coveredProduct Requirements Document (PRD)Tech StackDomain ModelingBlank CanvasFollow us online at thecodingagency.org.