Edith Harbaugh

Your experiments should live where your data already lives, and now they can! Warehouse-native Experimentation is now on BigQuery, Databricks, and Redshift alongside @Snowflake. Your warehouse stays the source of truth for everything. full details: bit.ly/3T2d108

If @jamescham goes to @saastr , I'd be honored to speak, but @jasonlk decides.

@jasonlk If a partner asked me to a partner meeting, I’d ask what they thought objections would be & the odds of investment given their firm.

It's totally OK to ask a VC at the end of a pitch meeting, "Based on what you know now, what are the odds you'd invest?" In fact, it can save a lot of time Just ask. "1-10, what are the odds you'd invest?" An even half-honest answer eliminates almost all of the ghosting anxiety ..

Agents break in production in all the ways you'd expect, and a few you wouldn't. 👀 Runtime Labs is all about staying in control when they do. Did we mention that no LaunchDarkly account is required to join? We'll see you there → bit.ly/4vRW1I9

There's a big difference between understanding runtime control and seeing it work in production. We're skipping the theory entirely and putting AgentControl in your hands. Save your spot → bit.ly/4vRW1I9

Awake before the coffee or bagel shops are open.

@karinklein 12/13 in my high school class who took the exam scored a 5 (including me!). I'd have to restudy if I was to take the exam tomorrow though.

@edith_h I love this. And you’re exactly the kind of person who proves the point.

“Calculus makes me feel smarter.” @karinklein advocating on the usefulness of advanced math. I took AP BC Calculus when I was 15 - don’t use it every single day but some of the concepts I do! #brainstormtech

Printing press also could have caused the Reformation. Will there be a similar revolt against AI slop?

“Just like printing press created more literature as it was easier to produce books, Claude will create more software & Applications.” @bcherny #brainstormtech

“To think about Claude productivity, have a shootout where one engineering team has Claude & another doesn’t - see which does better” @bcherny John Henry battle! en.wikipedia.org/wiki/John_Henr… #brainstormtech

@austencollins GP walked out of the room and didn’t return for ~20 minutes during the partner meeting for Series A. Afterwards, told partners he didn’t want to invest I hadn’t covered enterprise sales strategy. Which I had - when he had left. It was also in my physical deck.

@pitdesi Multimedia Gulch was also a thing, & then Social Gaming with Zygna en.wikipedia.org/wiki/Multimedi…

SF was over if not for OpenAI and Anthropic! Before that, it was over if not for Stripe, Uber, Airbnb, Twitter, Salesforce, Visa, Gap... and if you go back a bir further - Levi's, Bank of California, Spreckels Sugar, Wells Fargo, Union Iron Works, Southern Pacific :) The city gets “saved” by the next boom every time. After the gold rush, there was silver, railroads, shipping, banks, sugar, utilities, Pacific trade, defense, semis, PCs, biotech, enterprise software, the internet, social, mobile, SaaS, fintech, crypto, and AI. I think that's just how it goes in this town.

G2 named us a Leader in Feature Management for Summer 2026!!! Customer recognition means everything, so thank you for the trust and honest reviews. See the full report: bit.ly/4veRr6t

The teams pulling ahead have already figured out what comes after shipping AI. On June 11, @edith_h , Cameron Etezadi, and Marek Poliks are talking runtime control, production safety, and what hasn't worked. Come find out → bit.ly/3RmLB4o

Since launching in 2015, @LaunchDarkly has been synonymous with enterprise-grade feature management (50 trillion feature evals per day, serving many of the most exciting companies in the world). Today they're announcing AgentControl, the first product as they become the runtime control platform teams trust to move quickly with AI. While staying in control of code (and agents) in production, of course. Very excited for LD's next chapter! 🚀

@YouPulseX @JoshSteinVC That's what we have - the eval harness. launchdarkly.com/solutions/run-…

@JoshSteinVC @edith_h Git can prove the skill changed. Until the eval harness exists, what proves the PR comment got safer instead of just different?

Running my own little fleet of Claude agents — newsletter ingest, tweet drafts, code review on my PRs — has made one thing obvious: the prompt you shipped last week is often not the behavior running today. LaunchDarkly's AgentControl feels like the right shape of an answer. Excited to try it out!

It's an additional piece you want to have as you mature - you're decoupling deployment from release (in the words of the amazing @edith_h ) and allows you to be testing in production (h/t to @mipsytipsy ). That said, start by nailing your harness and quality gates (deterministic and adversarial). Then make sure your CI/CD is solid and you have excellent o11y so you know what's going on. Then adding feature flags for another level of testing makes all the sense in the world. I've not heard a lot about this in the dark factory world, but it's an obvious addition. Then it's time to add resilience engineering/game daying (I always think of @KoltonAndrus at Gremlin) to ensure that when your agentic swarm screws the pooch, it might take down checkout, but at least product pages still display or carts still work or whatever, There are fair reasons to argue for a different ordering, but that's a mature dark factory. as @nathenharvey and the DORA team keep pointing out, DevOps good practices matter more than they ever did.

@mattpocockuk Yes, we have customers using @LaunchDarkly for feature flagging with agents. We also have Agent Control to make it easier to control agents in production, after shipped. ld.team/6W88xI

Is anyone doing feature flag development with agents? Not tried it, but in theory feature flagging is an alternative model to PR's to getting work on main. 1. Put it on main, disabled by a flag 2. Deploy with the rest of the system 3. Unflag to selected users early 4. Fix bugs for those users 5. Unflag to more users 6. Repeat until shipped Feels like a perfect strategy to pair with agents

I started LaunchDarkly because teams deserved better control over what they shipped. With AI, that matters more than ever. Today we’re launching AgentControl: real-time visibility and control for AI agents in production. ld.team/6W88xI