Jesse Hall

Staff Developer Advocate @livekit 🥑 | YouTuber | Instructor | Opinions My Own ▶ 🦸🏿‍♀️ Become a 🦸🏼‍♀️

[Jesse Hall @codeSTACKr]

This week is my last one at @MongoDB.

4.5 years ago I joined as a Senior Developer Advocate with one clear mission: 👉 help developers understand MongoDB through video tutorials. I wanted to make complex database concepts actually make sense. That mission turned into hundreds of videos reaching millions of views, from the OG Jumpstart series debunking NoSQL myths to lounging on the beach showing off the MongoDB MCP server. Every video was a chance to break down another wall between developers and the tools they needed.

But some of the best teaching happened face to face. I traveled all over the world giving talks—some to small rooms, some to auditoriums with over 1000 people. The scale didn't matter. What mattered was talking to developers after the session, understanding what they were building, what problems they were stuck on, and how I could actually help. Those conversations shaped every piece of content I created.

The best part of this job was the partnerships. Working with teams at @vercel, @prisma, @mastra, @langchain, @DeepLearningAI, and many others taught me that great developer experiences happen when companies actually talk to each other. These collaborations pushed me to think bigger about what teaching could look like.

Notifications! **And this is going to be one of those awkward award speech moments where I just keep listing names**: @RitaMRDevIT, Lieke Boon, Pedro Machado, @nraboy, @anaiyaraisin, Tim Kelly, @LuceCarter1, @MeganGrant333, Ricardo Mello, @richmondalake, Apoorva Joshi, Pavel Duchovny, @BazeleyMikiko, Michael Lynn, Toni Bird, Dave Gruchacz, @mjasay, Dorothy McClelland, and so many others—you made every video better, every livestream less stressful, every event run smoothly, and every planning session actually fun. This team showed me what it looks like when people genuinely care about helping developers succeed.

Moving from Senior to Staff Developer Advocate meant getting to think more strategically about how we teach, who we partner with, and what content actually moves the needle for developers. That shift changed how I approach creating anything.

Friday is my last day. I'm taking everything I learned about teaching, community building, and developer advocacy into whatever comes next.

Now comes the worst part! I have to send back my laptop along with the stickers I've accumulated. 😩 I guess a clean slate is good. Send stickers please!! 😅

Thank you to everyone who watched a video, came to a talk, or spent five minutes at a MongoDB booth telling me about what you're building. That's why any of this mattered!!

[Jesse Hall @codeSTACKr]

@stolinski what did I miss here?

[Jesse Hall @codeSTACKr]

who will I see there??

[Cerebral Valley @cerebral_valley]

The second official Engineer World’s Fair Hackathon with @aiDotEngineer

Join us for 48 hours of building at the frontier of recursive self-improvement (RSI): systems that learn from their own outputs and become more capable through continuous iteration.

Excited to be partnering with @GoogleDeepMind, @MongoDB, @MiniMax_AI, @LiveKit, and @digitalocean.

🏆 $10,000+ in prizes and partner credits

June 27–28 hosted at @SHACK15sf

Apply below👇

[Jesse Hall @codeSTACKr]

I'm lovin supabase! so easy to work with.

[LiveKit @livekit]

Most voice AI agents forget you the second the call ends.

We built a starter kit that fixes it. LiveKit for the voice, @supabase for memory, RAG, and per-user auth, all in one Postgres database. TypeScript and Python included.

Give it a try and tell us what you're building.

[Jesse Hall @codeSTACKr]

@BhosalePratim I’ll be around. Let’s chat.

[Jesse Hall @codeSTACKr]

I’m in

[Thomas Dimson @turtlesoupy]

did you make it into the superintelligences? are you in the weights? https://t.co/2GE1WdVDRu

[LiveKit @livekit]

Voice AI has a benchmarking problem.

Everyone claims their end-of-turn model is the best, but you couldn't actually compare them. Datasets are proprietary, methods are opaque, and there is no shared ground truth.

That changes today.

We hit this while developing Turn Detector v1, so we open-sourced eot-bench. 5,000+ real user conversation turns across 14 languages, an evaluation harness that measures the real production tradeoff between latency and false cutoffs, and a live public leaderboard.

This should become the default way we evaluate turn detection models.

[Jesse Hall @codeSTACKr]

peak design

[Jesse Hall @codeSTACKr]

A voice agent that interrupts a user is the quickest way to frustrate them. Here's the fix 👇

[LiveKit @livekit]

We shipped LiveKit Turn Detector v1.

Instead of reading transcripts, it listens to speech directly, combining semantic and acoustic cues into one end-of-turn prediction.

The result: high accuracy, low latency—the best model we tested across 14 languages.

Available on LiveKit Cloud.

[Jesse Hall @codeSTACKr]

SpaceX is buying Cursor for $60 billion.

With xAI already in house, Musk now owns the model and the editor. One stack, top to bottom.

Cursor won by being model-agnostic. Hard to keep that up when the company upstairs makes Grok.

Let's see what happens to the default model.

[SpaceX @SpaceX]

SpaceX has exercised the option to acquire @cursor_ai in an all-stock transaction with the goal of building the world’s most useful AI models.

For the past few months, SpaceXAI has been jointly training a model with Cursor, which will be released in Cursor and Grok Build soon.

We look forward to working closely with the Cursor team to advance our frontier AI capabilities

[Jesse Hall @codeSTACKr]

@svpino It’s soooo good!
Have you tried it with LiveKit?

https://t.co/3byiz8n9xE

[Jesse Hall @codeSTACKr]

@BraydonCoyer @aarondfrancis So good!

[Jesse Hall @codeSTACKr]

I made a modified version of this called /improve-solo which uses all of the Solo orchestration features like todos, scratchpads, subagents, etc instead of local files.

Give it a try: npx skills add codeSTACKr/improve-solo

What do you think @aarondfrancis ?

[shadcn @shadcn]

You have Claude Fable for only a few days. Here's how to make the most of it.

Introducing /improve: use your most capable model to audit your codebase and write plans for cheaper models to execute later.

Studies your code, figures out bugs, perf, tech debt, missing tests, what to build and writes plans any agent can run.

[Aaron Francis @aarondfrancis]

Themes are coming!

[Jesse Hall @codeSTACKr]

I'm so sorry for my bad french..

[LiveKit @livekit]

We built a live multilingual, multi-person video call with Gemini 3.5 Live Translate on LiveKit. Everyone picks their language, speaks naturally, and hears each other in real time in their language of choice.

Watch the demo and check out the open source repo: https://t.co/xp8zU5NUqX

[Jesse Hall @codeSTACKr]

@shayneparlo @yacineMTB

[Jesse Hall @codeSTACKr]

@ErikCH Exactly

[Jesse Hall @codeSTACKr]

watching two very smart people argue about whether agents should write 100 percent of the code or 99 percent, when the percentage was never the interesting part.

nobody's losing sleep over an agent building an internal dashboard. the fear is it ships something that takes down prod at 2am.

so sort the work by how much it can hurt you. hand the agent the whole keyboard on the cheap, reversible stuff. stay in the loop where a bad diff actually costs you.

stop counting what percent is AI-written. ask what breaks if it's wrong. that's the job.

[Jesse Hall @codeSTACKr]

agi

[Anthropic @AnthropicAI]

Our internal data shows Claude is accelerating AI development—a possible path to recursive self-improvement, or AI autonomously building a more capable successor.

It’s happening faster than we thought, and the implications deserve greater attention. https://t.co/OVVPJO7VQx

[Jesse Hall @codeSTACKr]

Everyone's mad about Copilot's usage-based pricing. I think the meter is the best thing to happen to how people use these agents.

Completions are still free. What burns credits is firehosing 200k tokens of your repo at a frontier model and hoping. The runaway bills aren't a pricing problem, they're a discipline problem.

If you don't want a surprise bill, work tighter. Hand the agent the two or three files the task actually touches instead of the whole repo. Keep a cheap model as your default and only reach for a frontier one when the work needs it. Start a new chat for each task so you're not paying to drag an hour of history into every reply. And spend a little longer on the prompt before you fire it off.

You lose the predictable runway, sure. But a meter that makes you think before you spray context is a habit worth paying for.

[Jesse Hall @codeSTACKr]

@tdesseyn @AllThingsOpen And you!!

[Jesse Hall @codeSTACKr]

Look ma, I made it!!

So humbled to be among these amazing speakers.

I’ll be giving a 3 hour workshop on building open source voice AI agents for production at @AllThingsOpen.