That's Girl Code

For the season 3 finale, we're running a QBR (quarterly business review) on the podcast itself.

From the text that started it all through three seasons of struggles, wins, lots of learning, and two new jobs.

Then we dive into our goals season 4 live: more community, an in-person event, and a production upgrade.

We finish with our favorite and least favorite episodes, as well as reminisce on the memories that keep us going.

39 episodes in and we still love what we do!

See you in season 4!

In this solo episode, the last one of Season 3, Ellen picks up where Eden's MCP episode left off and asks how AI agents actually get to a company's data.

She starts with what an agent is, an AI that works toward a goal in steps and calls tools to reach things it can't touch on its own. She then walks through how software's access to data has changed: developers wrapping hand-written SQL in functions, then agents calling those same functions as tools, then MCP as the "menu" of tools any AI app can use.

From there she introduces Databricks and Genie. Genie began as a plain-English-to-SQL translator, and it has grown into your business analyst. The real challenge turned out to be giving AI the tribal knowledge about your data, not writing SQL. She breaks Genie into three pieces: Genie Agents that analysts teach, Genie One as the single front door, and Genie Ontology as the shared map of what your data means. She also covers how Genie reaches out to tools like Slack through MCP, and how it can become a tool inside someone else's agent, all while your data stays governed.

MCPs are everywhere right now. Can Eden explain them in 10 under 10 mins? Not sure, but she'll try.

In her last solo episode of Season 3, Eden breaks down the Model Context Protocol — the standard Anthropic released in 2024 and handed to the Linux Foundation in 2025 — using the hat business she and Ellen actually run as the example. What happens when you try to wire an AI agent to Shopify, Notion, Formspree, and Instagram DMs by hand? You build five custom integrations, maintain them forever, rewrite them when you switch models, and drown your context window in JSON. MCP is the fix.

In this episode:
• The problem MCPs solved (and why "the model is a brain with no limbs")• Hosts, clients, servers — what each piece actually does• Tools, resources, and prompts• The July 2026 spec change that made MCP stateless• The critiques: context bloat, the 67,000-token tax, and real security incidents

Eden and Ellen break down the economics of AI chips. They talk through how a chip actually gets made (wafers, logic dies, HBM memory), why one company in Taiwan (TSMC) manufactures nearly every advanced chip on Earth, and why NVIDIA's record-breaking quarter might be propped up by companies investing in each other in a circular loop.

90% of the Fortune 100 run their code through GitHub, and this week it went down for 7 hours. If your reaction to that is questions and confusion .... this one is for you.

Eden breaks down Git vs. GitHub (not the same thing!!): The 2005 drama that made Linus Torvalds rage-build Git in two weeks, the platform that built upon it and became a social network for developers, plus the bare minimum version control every vibecoder needs to know.

LLMs predict the next word. World models predict the next moment. Eden explains the AI "brains" powering Waymos and humanoid robots and Ellen gets into the real-world applications: simulated tornado testing, robot hands that can hold an egg, NVIDIA's Cosmos, and why people are getting paid to wear GoPros while doing chores. Your delivery robot is smarter than you think.

Ellen goes solo to tackle quantum computing!

She walks through quantum computing basics: classical bits vs. qubits, what superposition and the Bloch sphere actually mean, why rotating qubits creates wave interference that amplifies right answers and cancels wrong ones, why that gives quantum computers an exponential edge over brute force, and why all of it points toward Shor's algorithm eventually breaking RSA-2048 encryption.

Episode Description

In this episode, Eden and Ellen cover recent AI news.

1. Apple v. OpenAI LawsuitApple's federal lawsuit against OpenAI, two former employees, and hardware startup IO over alleged theft of trade secrets — plus the Musk-Altman fallout on X.

2. OpenAI Sandbox EscapeHow an OpenAI model being tested for hacking skills broke out of its sandbox and breached Hugging Face.

3. AI 2040 Scenario ForecastA look at AI 2040, a long-form projection of AI's trajectory through 2040, including proposals for international coordination and universal basic income.

4. AI and the Jacobian ConjectureHow AI disproved an 87-year-old math conjecture with a single counterexample.

Phia was supposed to be the Google Flights for fashion. An AI shopping assistant co-founded by Phoebe Gates and Sophia Kianni, backed by $43.5M and a Coachella lineup of celebrity investors. Then researcher Ben Edelman caught their browser extension cookie stuffing: silently taking affiliate credit for sales it didn't drive, possibly even swiping commissions from your favorite influencers' links. In this solo episode, Eden breaks down how the scheme worked (feature flags, background tabs, and a setting literally called "enable coupon auto drop"), why it only targeted iPhone Safari users, whether Phia's Series A investors were misled, and what past cookie-stuffing cases, from the eBay wire fraud convictions to the ongoing Honey lawsuits, might tell us about what happens next.

Sources mentioned in this episode:

• Ben Edelman's investigation: https://www.benedelman.org/phia-forced-clicks/
• Brian Dunning's Blog Post Alleging Ebay's Involvement in Cooking Stuffing: https://www.briandunning.com/fix_ebay.html

It’s the golden episode (season 3, episode 3), and Eden and Ellen are getting a little science-y. This one’s all about health tech wearables — the rings, watches, and bands tracking your every move, and what’s really happening under the hood.

We break down:

• How the hardware senses your body — optical PPG (photoplethysmography), ECG, LED light, and why the fingertip and earlobe are prime real estate for readings
• The metrics — heart rate, HRV, respiratory rate, blood oxygen, blood pressure, temperature, and how each one is actually derived
• Signal vs. noise — why your ring is more trustworthy asleep than mid-workout, and the adaptive filtering that cancels out your footsteps
• AI, machine learning & inference — the difference between measured data and educated guesses, and why you are your own baseline (and why that baseline is the real product lock-in)
• The competitors — Oura, Whoop, Garmin, Apple, and up-and-comers like the Lumia ear cuff
• Your data & who’s protecting it — privacy policies, consent screens, Washington’s My Health My Data Act, Senator Cassidy’s proposed HIPRA, and what happens to your data when a company changes hands

As always — we’re no doctors, lawyers, or geopolitical specialists. Stay safe out there.