AI Native Dev

Your weekly digest of AI software trends, expert insights, & handpicked content and events, delivered straight to your inbox.

[AI Native Dev @ainativedev]

Developers using AI tools are creating and merging twice as many pull requests — but AI-generated PRs have a 60/40 merge rate compared to 80/20 for humans. That gap reveals something important about how agents are actually being used in the wild: probing, experimenting, spawning throwaway work. Jellyfish's Nick Arcolano breaks down what the data actually says.

Watch the full episode at https://t.co/5BQXkP3uGn or listen wherever you get your podcasts

#AI
#agenticcoding
#claudecode
#codex
#AIskills

[AI Native Dev @ainativedev]

Most developers think scaling up means running more agents in parallel. But data from 250,000 real engineers shows there's a hard ceiling: even the most experienced developers max out at around 4 concurrent agents — and at that point, 80% of attention collapses onto just one. Nick Arcolano from Jellyfish explains why human attention is the actual bottleneck, and what it would take to truly break through.

Watch the full episode at https://t.co/5BQXkP3uGn or listen wherever you get your podcasts.

#AI
#agenticcoding
#claudecode
#codex
#AIskills

[Rohan Sharma @rrs00179]

Day 2 of AI Native DevCon

• Context needs ownership, not just documentation
• Agent behavior needs measurement, not vibes
• Safe execution boundaries matter
• Skills scale better than giant prompts
• Agent enablement needs ownership

Full Day 2 recap
https://t.co/zeKjVg0lI6

[AI Native Dev @ainativedev]

AI is making engineering teams faster — but faster at what, exactly? We've entered a world where companies can hire infinite robot contractors, but most organisations weren't designed to reason about infinite capacity. Nick Arcolano's "jet engines in cars" analogy captures the core tension of 2026: the bottleneck isn't code generation anymore. It's everything around it.

Watch the full episode at https://t.co/5BQXkP3uGn or listen wherever you get your podcasts

#AI
#agenticcoding
#claudecode
#codex
#AIskills

[AI Native Dev @ainativedev]

@Shashikant86 @evalstate damn!! no worries, recordings will be there.

[AI Native Dev @ainativedev]

As we wrap up two fantastic days at AI Native DevCon, Birgitta Böckeler, Distinguished Engineer at @thoughtworks delivered a keynote focused on the bigger picture.

One of the key themes was understanding the different roles played by the model, the coding harness, and the coder harness in modern software development.

A great reminder that successful AI-assisted engineering is about much more than model capability alone.

Thank you to our speakers, attendees, and everyone who joined the journey.

[Liran Tal @liran_tal]

Closing keynote with Birgitta Bockeler from @thoughtworks about the state of AI coding assistants at @ainativedev 🪇

#ainativedevcon

[AI Native Dev @ainativedev]

Former @Docker CTO Justin Cormack (@justincormack) shared a powerful reminder today:

Passing tests doesn't always mean your system is correct.

As AI-generated software becomes more common, observability is emerging as a critical tool for validating behavior, enforcing security requirements, and building confidence in what agents produce.

Great session on the realities of AI-native engineering.

[AI Native Dev @ainativedev]

Ian Thomas from @Meta is exploring what AI Native Engineering looks like in practice.

Not just better coding tools, but a shift in how engineers spend their time, how teams measure impact, and how organizations adapt to a world where agents are part of the workflow.

The future of software engineering is being written now.

[AI Native Dev @ainativedev]

Packed workshop with Aashrey Tiku and Lamis Mukta (@lamismukta) from @AnthropicAI.

Attendees are getting hands-on experience with Claude Managed Agents, building and deploying a working agent while learning the infrastructure patterns behind production-ready agent systems.

A great look at what it takes to move from prototype to deployment.

[AI Native Dev @ainativedev]

Why Developers Hit a Wall at 4 AI Agents

Engineering teams are doubling their pull request output with AI — but the data shows that nearly 40% of AI-generated PRs never get merged. Nick Arcolano from Jellyfish tracks real agentic coding behaviour across 250,000 developers, and what he's finding challenges some of the most common assumptions about productivity, multi-agent scaling, and what it means to justify AI spend to your CFO. This isn't sentiment data or anecdote — it's the actual signal from the industry.

Watch the full episode at https://t.co/5BQXkP3uGn or listen wherever you get your podcasts

[Liran Tal @liran_tal]

Amit Kushwaha starts off immediately diving into the tech behind tokens and goes next to explain KV caching #ainativedevcon

The importance of hitting the cache and the ability for the LLM server to prefill the tokens quickly (time to first token)

[AI Native Dev @ainativedev]

Packed room for Derek Ashmore's (@Derek_Ashmore) workshop, "The AI Agent Testing Pyramid."

Everyone wants to build AI agents. Fewer teams know how to test them.

Today's workshop is exploring practical approaches to validating non-deterministic systems, from deterministic testing patterns to LLM-based evaluation.

The workshop repo is public if you'd like to follow along:

https://t.co/I6zaze7LK0

[Rohan Sharma @rrs00179]

Day 1 of AI Native DevCon

• Skills are becoming software assets
• Context is infrastructure
• Verification is the new bottleneck
• Security can't be bolted on later
• Agent adoption is a team challenge, not just a tooling one

Full Day 1 recap 👇
https://t.co/VceLtL7hfO

[AI Native Dev @ainativedev]

Day 2 of DevCon is shaping up to be another great day.

We've highlighted a few Tessl (@tessl_io) sessions worth adding to your schedule :
• Harness Engineering Beyond Code with Marc Sloan
• From Vibes to Metrics with Simon Obstbaum & Rob Willoughby
• Don't Write Prompts, Write Software with Baruch Sadogursky (@jbaruch) & Macey Baker (@macebake)

Plus sessions from Ian Thomas (@Meta), Shaun Smith (@huggingface), Katie Roberts (@nearform), and other leaders helping define the future of AI-native engineering.

Swipe through and start planning your day.

Visit https://t.co/Kl88dtYu4R for more.

[AI Native Dev @ainativedev]

@lamismukta 🫀

[AI Native Dev @ainativedev]

Dave Farley (@davefarley77) has just taken the stage at AI Native DevCon.

As the founder of Modern Software Engineering and creator of a YouTube channel followed by more than 260,000 software engineers, Dave has spent years helping teams build better software.

Now he's tackling one of the biggest questions in our industry:

Is vibe coding really the best we can do?

Join us as Dave explores what AI changes, what still matters, and what programming might look like in the years ahead.

Watch live:
https://t.co/W7a8kHJ7Ix

Register now for virtual session to stay updated:
https://t.co/uU8vvK0hux

[AI Native Dev @ainativedev]

Lamis Mukta (@lamismukta), Member of Technical Staff at @AnthropicAI, has just taken the stage at AI Native DevCon.

Memory has become a core building block for AI agents.

But what happens when memory isn't enough?

Lamis is exploring a fascinating idea: dreaming. A process that helps agents curate, prune, and learn from their memories to improve performance over time.

Watch live:
https://t.co/W7a8kHJ7Ix

Register now for virtual session to stay updated:
https://t.co/uU8vvK0hux

[AI Native Dev @ainativedev]

Ryan Lopopolo (@_lopopolo ), Member of Technical Staff at @OpenAI, has just taken the stage at AI Native DevCon.

Most teams are still using AI with workflows designed for a world where humans write all the code.

Ryan is exploring a different model.

Humans steer through goals, constraints, and feedback loops. Agents execute.

If you're trying to move from AI coding demos to real production workflows, this is one to watch.

Watch live:
https://t.co/W7a8kHJ7Ix

Register now for the virtual session to stay updated:
https://t.co/uU8vvK0hux

[AI Native Dev @ainativedev]

@altryne @liran_tal @_lopopolo @OpenAIDevs it will be shared soon!