AI Native Dev

The best part of AI DevCon NYC wasn't on stage. It was the hallway chats, the workshops, and people swapping what's working. Thank you, NYC. Next up London, June 7-8: luma.com/aidevcon-ldn2027

Agentic coding isn't making software worse, according to Maggie Appleton of GitHub Next. Beyond the obvious slop, she sees bug reports falling and shipping speeding up at big companies. The execs demanding raw speed, she argues, will soon point agent tokens at quality instead. Watch the full episode at youtu.be/0cySwRF7oJI or listen wherever you get your podcasts

Full house for one of the most practical sessions of the week. Peyton Montei and @jkcso showed how trails, per-commit evals and shared memories keep agent decisions from getting lost between sessions. Get your London tickets, June 7-8: luma.com/aidevcon-ldn2027

Spec-driven development is where Maggie Appleton of GitHub Next thinks we're inevitably heading. If tests verify the code does what it says, how much is anyone really going to read it? She argues the future is reviewing high-level plans and agent reports, not line-by-line diffs. Watch the full episode at youtu.be/0cySwRF7oJI or listen wherever you get your podcasts

Looking back at AI DevCon NYC: what if your agent's code had to prove it was right first? Nada Amin, Associate Professor at Harvard, and Fernanda Graciolli of Midspiral showed how: write the contract, let the agent iterate, ship what the verifier accepts. See you in London, June 7-8: luma.com/aidevcon-ldn2027

3 days in Brooklyn. Here's AI DevCon NYC in one video. Workshops, talks on what breaks when you scale coding agents, a packed sponsor floor, and a party to close it out. Thank you, NYC. See you in London, June 7-8: luma.com/aidevcon-ldn2027

That's a wrap on AI DevCon NYC. 3 days in Brooklyn: a workshop day, two days of talks, and a lot of honest conversations about what it takes to ship with coding agents. See you in London, June 7-8, 2027: luma.com/aidevcon-ldn2027

Managing one agent after another all day? @JackWoth98 from @GoogleDeepMind on moving from a local coding agent to remote managed agents that verify their own work in loops you can step out of. When it comes back, it’s done.

Thank you to the sponsors who made AI DevCon NYC happen. @MLHacks @Bluebear @launchdarkly @snyksec @CockroachDB @tweetsbyport @dash0hq @codeplain From the Agent Genius Bar to the workshops to the party, you made this week what it was. And thank you to our partners DevIT, Bond AI, The AI Summit New York, and Tech Impact for helping bring this community together.

Day 3 of AI DevCon NYC and the great guests keep coming. @jaydestro from Microsoft on why you shouldn’t trust an agent without real context. Test what it writes, then improve what it works from, from docs to MCP servers.

Coding agents make every developer faster, but Maggie Appleton of GitHub Next argues that speed alone doesn't add up to a product. When everyone races ahead on their own branch, you get scary merge conflicts and features nobody agreed to. As she puts it, nine women can't make a baby in one month. Watch the full episode at youtu.be/0cySwRF7oJI or listen wherever you get your podcasts

That’s day 2 of AI DevCon NYC done. Thank you to today’s speakers and everyone in the room and online. You brought the hard questions and the honest war stories. We’re off to the after party now, and we’ll be back tomorrow for more.

We went from 6 laptop factories to 1 shared pipeline. Then PRs piled up, agents fought over the same files, every merge forced a rebase, and CI costs climbed. @patrickdebois on what breaks when a whole org shares a software factory, and what teams are trying: splitting work by code region, speculative merge queues, semantic merge checks, and human sign-off on risky paths. Before you open the factory to everyone, get tests, observability, and contract rules in place.

Agentic coding still runs through a text box, and Maggie Appleton of GitHub Next says that very textual, simple interface isn't cutting it. She wants coding agents that research ahead of the plan and prompt you with decisions, instead of waiting to be prompted. It's a different picture of how humans and agents should work together. Watch the full episode at youtu.be/0cySwRF7oJI or listen wherever you get your podcasts

Prompting an agent not to force push is a suggestion. Removing force push from Git is a guarantee. @GeoffreyHuntley at AI DevCon NYC: one Nix file defines your laptop, CI, agent sandbox and production. Patch any dependency once, transitive ones included, and it applies everywhere.

Day 2 of AI DevCon NYC has started in Brooklyn, and it kicked off to a full room. @guypod opened with the welcome keynote and gave everyone a frame for the day: we’re moving from building the car to building the factory. Skills define the work, loops run it and learn from it, and factories tie it all together. So, devs: which part of your workflow would you hand to a loop first? Code review, security fixes, tests? Not in the room? Grab a virtual ticket and watch along: aidevcon.io

A software factory doesn’t arrive in one leap. Today at AI DevCon NYC, a full room took the first steps. Tomorrow at 9am ET, Guy Podjarny on building a software factory you can trust: aidevcon.io

Sorry @LogRocket, our sponsor slots sold out 😅 Respect for the hustle though. Best parking-spot booth at AI DevCon NYC 🚗 Better luck next year, we’ll save you a table (indoors). If you're a sponsor interested in a spot for next year, get in touch!