Alem Tuzlak maintains the devtools half the React ecosystem depends on: TanStack AI, TanStack DevTools, and the devtools for Remix and React Router. He builds AI tooling for a living at CopilotKit. And he thinks the newest AI models are getting worse.

We talk about why every Claude release since Opus 4.5 has felt like a downgrade to him, the experiment where two coding agents worked 30 hours and finished 6 of 20 tasks while a third finished the rest in 6, and the creepier problem underneath it all: the AI that gave his doctor wife a wrong answer, got corrected, and replied "you're right, I was supposed to tell you this." We also get into what makes someone a developer when anyone can subscribe for $200 a month, why that $200 is half the minimum wage in his country, how he runs six agents at once without losing his mind, and why he gave up on coding tutorials entirely.

Chapters

00:00 — Welcome and how a grandfather's memory led to open source 01:55 — Call of Duty 4, design skills, and becoming a developer 04:55 — "I'm gonna lose my job": first impressions of AI and what it did to outsourcing 07:00 — Agent software factories, the OpenCode hype cycle, and what excites him in 2026 09:05 — Maintaining TanStack AI: community vs roadmap, and AI slop PRs 11:50 — Generative UI, Jarvis, and CopilotKit's bet 17:00 — What separates a developer from a $200 subscription 18:00 — The ChatGPT diet plan that was wrong on purpose 20:15 — His wife is a doctor. The AI said "I was supposed to tell you this." 22:40 — How TanStack AI is architected: middleware everywhere and talking to 20 agent frameworks 27:30 — Forge 42, Code Forge, fatherhood, and the story behind the "I need a job" tweet 30:10 — The 30-hour experiment: Claude Code and Codex vs the new contender 35:15 — Why Opus 4.5 was the peak: downgrades and the intelligence plateau 37:30 — His actual workflow: the grill-me skill, writing his own skills, and Kira.dev 41:10 — Local AI and the parts of the world a $200 plan leaves behind 45:10 — Coding tutorials are dead (the watch-time data says so) 47:20 — Nobody knows about your project? It doesn't matter: marketing for maintainers 49:15 — New in TanStack AI: WebSockets, generic interrupts, sandbox snapshotting 51:05 — AI content, putting a face on it, and the slop economy 55:20 — If everything he built disappeared: his 30-second message to the world

Are we ready for the future of AI?

AI is changing how we build software, how developers work, and potentially how entire careers evolve. But how much of what we hear about AI is actually true, and what should developers and builders be doing right now to prepare?

In episode 3 of Talking Tech, we sit down with Zahwah Jameel, a builder advocacy community manager, to talk about the rapidly changing world of AI, developers, and technology.

We discuss the reality behind AI replacing jobs, the rise of AI agents, how developers are adapting to AI-assisted coding, and why the future may belong to people who know how to work with AI rather than simply compete against it.

Zahwah also shares her journey from writing and coding in college to technical writing, working with startups, and eventually moving into developer advocacy and builder communities.

Guest: Zahwah Jameel

Chapters

00:00 Cold open: are we scared of AI?

00:25 Intro

00:31 Welcome, Zahwah

01:00 The future of technology and AI

01:50 What nobody has figured out about AI

03:00 From developer advocacy to builder advocacy

04:10 What does a builder advocacy manager do?

06:00 Working with developer communities

08:00 Zahwah's journey into tech

10:00 From writing and coding to technical writing

12:15 Working with startups

15:00 Building for developers

18:00 From technical writer to developer advocate

21:00 Getting into developer advocacy

24:00 What developers actually need

27:00 How AI is changing developer work

30:00 AI, jobs and the future of work

33:00 The rise of AI agents

36:00 What happens when AI agents become normal?

39:00 Should developers be worried about AI?

42:00 Learning to work with AI

45:00 AI-assisted coding and development

48:00 Writing for humans vs writing for AI

51:00 The future of technical writing

54:00 Building skills in the AI era

57:00 Advice for developers and builders

1:01:00 What should you learn next?

1:05:00 The future of AI and technology

1:09:00 Final thoughts

1:11:38 Outro

What does it take to break into tech with no background, no connections, and no safety net?

Paul Bratslavsky went from refugee to Brazilian jiu-jitsu instructor to software developer, landing his first coding job at 39 for $13.75 an hour. In this episode of Talking Tech, we get into the mindset behind that journey, what AI is actually doing to the developer job market, why most tech content sounds identical right now, and how to stand out when everyone has the same tools.

This one went long for a reason.

Guest: Paul Bratslavsky, youtube.com/@codingwithpaul

Chapters

00:00 Cold open

00:13 Intro

00:19 Who Paul is: refugee, coding at 35, first job at 39

02:17 What he's building now: a local AI video editor

03:40 Shipping products in days

04:27 From "how do I build it" to "what problem do I solve"

05:53 Juniors, seniors, and the shrinking entry path

06:29 Proving AI saves time, not token maxing

07:19 Token pricing: $100 in two hours

09:02 When convenience beats understanding

10:10 Has the path for junior developers changed?

11:23 Your value moves from building to supporting

13:23 Spec files and knowing what you want

14:27 Build what AI can't one-shot: BM25 and vector search

16:31 How much does a developer's brand matter?

17:45 How a Udemy course led to a job at Strapi

20:31 Become the known expert in one tool

23:14 What running a jiu-jitsu school taught him

24:12 Asking for free groceries

25:26 Opening a school as a blue belt

26:38 Excuses, grifters, and burning the bridges

29:47 "You already have a business": one thing at a time

30:17 Leaving the jiu-jitsu world behind

31:42 Moving into developer relations

35:21 Pick a niche: the friend who only builds payment gateways

37:01 You can't market to developers, you help them

39:25 The hardest part of devrel content

44:10 Finding tools through Claude instead of Google

48:37 How a dev tool gets recommended by an LLM

51:57 Standing out when everyone sounds the same

55:08 The guitar short, and becoming a student of attention

57:53 AI-made work that isn't slop

59:35 YC's letter: use AI, don't sell AI

1:01:08 The fish parable: subsidized AI and token pricing

1:04:54 Dev pushes back: paying for speed is still worth it

1:08:39 Why founders and CEOs are bullish on AI

1:10:31 What jobs look like next

1:13:12 Replacing his subscriptions with Claude

1:14:44 Selling the service, not the software

1:16:24 Hedge yourself: get a job or ship a product

1:18:16 Paul's next venture: music meets tech

1:21:06 The 30-second message: everything is borrowed

1:23:11 Wrap-up

1:23:47 Outro

Karl Hughes runs Draft.dev, a content agency that writes for developer-tool companies, and The Podcast Consultant, a podcast production business he bought. He was the first engineer at three startups, started Draft.dev in the first months of COVID, and now buys small service businesses on the side.

We talk about why he tracks every hour of his own time, his rule of only doing the things only Karl can do, taking a full month off from the business in year two, what a service agency actually sells for, why his ships are always on fire and which fires he lets burn, product-market-founder fit, whether you can build a dev-tool business without living in San Francisco, what AI did to the price of a blog post, and why consistency over 30 years beats three hard years.

Guest: Karl Hughes, draft.dev

Chapters

00:00 Cold open

00:14 Intro

00:20 Who Karl is and what Draft.dev does

02:20 Origin story: first engineer at three startups

03:30 2020: the startup falls apart and Draft.dev takes off

04:50 When he knew it had legs: half his time on new business

06:37 Why he tracks every hour of his own time

07:07 From writing software to writing about it

08:11 Skill stacking: decent at three things beats best at one

08:59 Running two companies on 20 hours of real work

11:19 The month-off test: a business that runs without you

12:44 "If I'm not in my business, what's my leverage?"

14:26 Product doesn't matter, market penetration does

15:33 Entrepreneurship through acquisition

17:37 What a service business actually sells for

19:34 The ships are always on fire: which problems he ignores

21:32 Buying a company that's on fire

24:03 Entering a niche you know nothing about

26:47 Do you have to be in San Francisco?

28:58 Product-market-founder fit

30:19 His unfair advantage

33:32 GEO and AEO: nobody has the answer

37:25 Selling something you've never done before

41:53 Why dev tools is a hard industry to serve

43:48 What AI did to the price of content

45:37 If content is infinite, what is it worth?

47:04 AI adoption is further behind than you think

50:06 Opinion and personality as the last leverage

52:39 Buying back your time and the levels of entrepreneurship

54:05 Mindset, cutting out cynics, finding a bigger room

56:59 Manifestation: belief, action, results

59:48 When to pivot: time-box the decision

1:02:05 "I'd have a million-dollar business in a year"

1:03:07 The 30-second message: consistency over 30 years

1:04:30 Wrap-up

1:05:02 Outro