The Programming Podcast

This isn't about a new JavaScript framework; it's about a significant shift in software engineering. If you're in tech or aspire to be a software developer, this discussion is crucial for understanding the future of web development.

✅ If you made it to the end, comment: “Claude Code!” so we know who is a real one!

In this episode, Leon and I live-react to the viral tweet from Andrej Karpathy (founding member of OpenAI, former Tesla AI Director) that has the entire dev world talking. Karpathy reveals that his workflow has shifted from 80% manual to 80% agentic in just one month. We’re diving deep into what it actually means to "program in English" and whether our skills are atrophying as AI takes the wheel.

SITE https://www.programmingpodcast.com/

💡 Sponsor: Level Up Financial Planning
Changing careers or increasing your income? Get financial clarity with Level Up Financial Planning—helping early and mid-career tech professionals secure their financial future. Visit LevelUpFinancialPlanning.com for a free consultation!
https://www.levelupfinancialplanning.com/

Stay in Touch:
📧 Have ideas or questions for the show? Or are you a business that wants to talk business?
Email us at [email protected]!

Danny Thompson
https://x.com/DThompsonDev
https://www.linkedin.com/in/DThompsonDev
www.DThompsonDev.com

Leon Noel
https://x.com/leonnoel
https://www.linkedin.com/in/leonnoel/
https://100devs.org/

📧 Have ideas or questions for the show? Or are you a business that wants to talk business?
Email us at [email protected]!

We also discuss:

- The "Vibe Coding" Shift: Why the "Idea Guy" is dead and why the "Builder" is winning.
- Atrophy vs. Expansion: Are we losing our edge, or just building faster?
- The "Slop-ocalypse": Preparing for 2026 as GitHub gets flooded with AI-generated spaghetti code.
- Real-world Impact: How I’m using these exact agentic workflows to build EventConvo.

Special Guest Mention: Shout out to Logan Kilpatrick from the Google Gemini team for the insights that sparked this conversation.

Chapters
0:00 – The Biggest Change to Coding in 20 Years
1:05 – Shoutout to Logan Kilpatrick & Gemini
1:52 – Reacting to Andrej Karpathy’s Viral Tweet
2:40 – "I am Programming in English Now"
4:05 – How Our Prompting Has Evolved (Whisper Flow & Opus)
5:30 – Does Agentic Coding Feel "Wrong"?
6:55 – The Theo T3 "Toothbrush" Workflow
7:45 – Building EventConvo: AI as a Force Multiplier
9:15 – Is AI Taking Developer Jobs? (The Stack Overflow Comparison)
12:35 – The Death of the "Idea Guy"
14:15 – Why You Still Need an IDE (Watching Agents Like a Hawk)
18:30 – The Senior Dev Gap: Pattern Recognition vs. Logic
20:20 – The Fear of Skill Atrophy
23:10 – Speed Ups vs. Expansion: Are We Doing Too Much?
27:40 – Leverage: Stop Telling AI What to Do
32:10 – Is Coding Still Fun?
35:50 – 2026: The Year of the "Slop-ocalypse"
43:05 – What Happens to the 10x Engineer?
48:20 – Future Tech: Anthropic Co-work & Gemini CLI
52:45 – Ask Danny & Leon: How to Crush a 3-Hour Virtual Onsite
59:20 – Final Thoughts: Now is the Time to Play

Apple’s latest decision to team with Google and use Gemini signals something bigger than a headline. It highlights how far Gemini has come, and why utility is becoming the real battleground for AI in 2026.

In this episode, we sit down with Logan Kilpatrick from the Gemini team at Google DeepMind to talk about what this moment means for developers, builders, and teams choosing tools today. We get into the evolution of Gemini, why multimodality and long context mattered early, and how Gemini is increasingly showing up where people already work through products like Workspace, Gmail, and Docs.

✅ If you made it to the end, comment: “Documentation Traffic.” so we know who is a real one!

SITE https://www.programmingpodcast.com/

Stay in Touch:
📧 Have questions for the show? Or are you a business that wants to talk business?
Email us at [email protected]!

Danny Thompson
https://x.com/DThompsonDev
/ dthompsondev
www.DThompsonDev.com

Leon Noel
https://x.com/leonnoel
/ leonnoel
https://100devs.org/

📧 Have questions for the show? Or are you a business that wants to talk business?
Email us at [email protected]!

Subscribe if you want honest, no-hype conversations about where software development is heading in 2026.

We also dig into what’s next for image generation and editing, including the leap in text accuracy, practical features creators want (like layers and transparent backgrounds), and the role of technologies like SynthID in building trust as AI content becomes more common.

Plus: vibe coding, reasoning models, prompt expansion, how developers can ask for more ambitious outcomes, and why proactivity and proof-of-work experiences are shaping the next wave of AI products.

If you’re building with AI or trying to stay ahead as a developer, this episode will give you a clear view into where things are going.

0:00 Apple’s decision changes the AI narrative
1:11 Meet today’s guest: Logan Kilpatrick (Gemini team)
1:52 Apple quote and why this is big for Gemini
2:19 Are the model wars over, and are utility wars next?
4:20 Gemini in real workflows: Workspace, Gmail, and daily utility
5:26 AI hype vs reality: why products must meet users where they are
6:42 Balancing frontier progress with “do not break the product” responsibility
8:25 Nano Banana and the leap in text accurate images
9:28 Infographics, presentations, and where image generation is going
10:00 Why text rendering became a key metric for model quality
12:19 What users want next: transparent backgrounds, layers, better editing
13:52 Trust and safety: SynthID, watermarking, and standardization challenges
16:06 What developers ask for most: quota and scale
17:38 Logs and the idea of an agent that finds missing product features
19:31 How people should use AI differently: native form factors, proactivity
21:19 Proof of work and why Deep Research builds trust fast
24:53 Prompt expansion, reasoning models, and why “prompt wizardry” is fading
26:40 Ask for more: why ambition is now the best prompting strategy
29:36 Frameworks vs over optimizing prompts: when it matters, when it does not
32:52 What “vibe coding” means now and how it is evolving
34:08 Teaching vibe coding on day one of school
36:01 Ask Danny, Leon, and Logan: feeling behind in tech
39:02 The moving finish line, authenticity, and turning the gap into a roadmap
41:25 Closing

Is Tailwind actually “dying” or are we watching something bigger happen in real time?

In this episode of The Programming Podcast, we dig into a brutal 2026 reality check for developers: the tools and communities we’ve relied on for years are getting shaken up by AI in ways most people didn’t see coming.

We break down the Tailwind situation, including the GitHub discussion that surfaced a behind-the-curtain truth about modern open source: documentation traffic is the funnel, and AI is bypassing it. When LLMs stop sending humans to your docs, what happens to paid products, sponsorships, and ultimately the ability to maintain the framework at all?

Then we shift to the other major signal: Stack Overflow’s decline. The charts are ugly, and the reasons are deeper than just “people use AI now.” We talk about the human cost of hostile community dynamics, why developer Q&A moved to Discord long before LLMs took off, and what it means when the internet’s most “verified” coding resource starts fading.

Finally, we go big-picture: if open source loses funding and loses maintainers, what happens to the future of software? And where do new developers go when both the old learning paths and the new AI-heavy paths are unreliable?

✅ If you made it to the end, comment: “Documentation Traffic.” so we know who is a real one!

SITE https://www.programmingpodcast.com/

Stay in Touch:
📧 Have questions for the show? Or are you a business that wants to talk business?
Email us at [email protected]!

Danny Thompson
https://x.com/DThompsonDev
/ dthompsondev
www.DThompsonDev.com

Leon Noel
https://x.com/leonnoel
/ leonnoel
https://100devs.org/

📧 Have questions for the show? Or are you a business that wants to talk business?
Email us at [email protected]!

Subscribe if you want honest, no-hype conversations about where software development is heading in 2026.

It’s 2026. If you want to learn how to code and actually get results, this episode is your roadmap.

Danny and Leon break down the stuff that is not “cool advice,” but is the advice that gets people hired. We talk about why course-hoarding feels productive but is not, what the real bottleneck in tech actually is (hint: not syntax), and how to build a learning system that survives the trough of sorrow.

You’ll walk away with:

The 3-part foundation that decides if you stick with coding

Why “problem solving” is the real job, not typing code

When AI helps, and when it quietly ruins your learning

The learning science that separates dabblers from finishers (active recall + spaced repetition)

How to pick what to learn based on your local market (and stop wasting years)

Why community is the cheat code that keeps you moving when motivation dies

If your goal is to learn to code this year, do one action right now: join a community, pick a path, and start.

✅ If you made it to the end, comment: “In 2026 I’m gonna grow my skill set.”

SITE https://www.programmingpodcast.com/

Stay in Touch:
📧 Have questions for the show? Or are you a business that wants to talk business?
Email us at [email protected]!

Danny Thompson
https://x.com/DThompsonDev
/ dthompsondev
www.DThompsonDev.com

Leon Noel
https://x.com/leonnoel
/ leonnoel
https://100devs.org/

📧 Have questions for the show? Or are you a business that wants to talk business?
Email us at [email protected]!

Chapters

00:00 - The 2026 promise: real results, not cool advice
01:56 - Why you should listen to us (and why this episode goes viral)
04:18 - Start with your “why” or you will quit
05:45 - The 3 foundations: manage frustration, consistency, take care of yourself
09:11 - What companies pay for: problem solving, not syntax
12:30 - The shiny becomes dull: how people quit when novelty fades
15:49 - Build to derisk yourself: projects that help vs projects that hurt
17:15 - The trough of sorrow and crash of ineptitude
18:51 - AI as the easy exit, and why you cannot take it
21:09 - Hot take: earn AI later, do not use it early
22:46 - Over-reliance makes you weaker (and why teams sometimes remove AI)
25:36 - Learn how to learn: Barbara Oakley and rebuilding your foundation
26:45 - Active recall: do less work, remember more
27:45 - Spaced repetition: stop forgetting what you learn
30:11 - Build projects early to create incentive and momentum
32:25 - Make concepts “sticky” with analogies and mental models
33:31 - Your first year on the job decides career vs job
35:37 - Stop asking for “the tech stack” and start reading your local market
38:40 - Dallas example: 400 MERN jobs vs 34,000 Java jobs
42:31 - No one is taking risks right now: prove you can hit the ground running
44:16 - Target companies strategically (emerging talent programs, hiring volume)
47:00 - Do not marry a stack, marry adaptability
48:49 - System rules: make every project your own, break goals into daily actions
51:38 - Pick one path and finish: FreeCodeCamp, Odin Project, 100Devs, Full Stack Open
52:42 - The real cheat code: community
56:31 - Motivation expires, community keeps you moving
59:44 - Talent is distributed, opportunity is not
1:03:28 - Take action right now: join a community
1:04:34 - Comment to prove you are a real one
1:04:56 - “I don’t have time” and the priority truth
1:06:10 - Your value in 2026: caring, context, judgment, and composing solutions

The “prompt-and-pray” era is over — and that’s a good thing.

In this episode, we break down why AI “magic” collapses under real production traffic (edge cases, hallucinations, messy inputs, and even infrastructure-level failures)… and what replaces it: actual AI engineering.

Danny frames the shift with four architectural pillars that make LLM features shippable and reliable:
- State orchestration (stop treating models like employees — they’re stateless CPUs)
- Constraint generation (JSON forcing, schema-driven outputs, type-safe sampling)
- Infrastructure reliability (retries, backoff, fallbacks — because inference can and will fail)
- Regression testing & evals (measure prompts like code, break builds when quality drops)

SITE https://www.programmingpodcast.com/

Stay in Touch:
📧 Have questions for the show? Or are you a business that wants to talk business?
Email us at [email protected]!

Danny Thompson
https://x.com/DThompsonDev
/ dthompsondev
www.DThompsonDev.com

Leon Noel
https://x.com/leonnoel
/ leonnoel
https://100devs.org/

📧 Have questions for the show? Or are you a business that wants to talk business?
Email us at [email protected]!

We also hit the reality of agent “throughput” vs human review bottlenecks (Phoenix Project vibes), why monolithic agents are a trap, and a listener question about networking + credibility after pitching an MVP that isn’t fully shipped yet.

If you’re building AI features for real users — not demos — this is the blueprint.

00:00 — The “prompt-and-pray” era is over
02:49 — AI hype fades: guardrails + reality
06:34 — Deterministic software vs probabilistic models
07:29 — The 4 pillars of AI engineering (overview)
11:37 — Pillar 1: state orchestration (FSM, stateless models)
20:26 — Pillar 2: constraint generation (JSON, schemas, type safety)
28:28 — Pillar 3: infra reliability (retries, fallbacks, failures)
32:21 — Pillar 4: evals + regression testing (LLM-as-judge)
43:40 — Listener question: networking, MVP pressure, and credibility