Leo Jiang

AI @ Amplitude. Ex FB, Google. I build cool stuff for fun

[Leo Jiang @_leojiang]

@ashleybchae @jaeyun_ha Amplitude's looking for founders to build & lead 0-1 products. The idea of training a browser use agent has been floating around!

[Leo Jiang @_leojiang]

And I get a nice summary:

[Leo Jiang @_leojiang]

Done this a few times, surprises me every time:

I tell Cursor/Claude Code to log timings and improve performance by 50%, and it just... does it.

Doesn't just work for AI slop, it worked on my hand-written pre-Cursor code I spent a lot of time optimizing.

[Leo Jiang @_leojiang]

@spenserskates It's cool to view the topics and prompts for all the brands in one place, e.g. https://t.co/uijv7rxF2b

[Leo Jiang @_leojiang]

Anyone else finding that Composer 2.5 is pretty bad at following Cursor rules? At least it's fast at generating linter rules to replace Cursor rules

[Leo Jiang @_leojiang]

Really liking Composer 2.5, feels really fast, at least twice the speed of Sonnet 4.6/GPT 5.5

[Leo Jiang @_leojiang]

SEO used to be about ranking pages. Now it’s about becoming the source AI systems cite. Controlling the narrative around your brand is critical for getting AI to recommend your products. I built Amplitude's AI Visibility product and we've been using it internally to improve our own performance in ChatGPT, Google AI Overviews, Claude, and Gemini.

Since mid-2025, we began noticing a pattern: competitors were publishing large volumes of AI-generated content highlighting their benefits while mentioning outdated or inaccurate information about Amplitude. When ChatGPT answered questions about Amplitude, it would often cite our competitors' content and produce heavily biased or incorrect results.
For a while, our strategy was simple: continue publishing high-quality, human-written content. We assumed Google and OpenAI would quickly filter this type of content, but those AI articles are still being cited today.
So we shifted our strategy and started publishing AI-assisted content as well, but with strict quality controls:
- human review on every article
- accurate product context from our internal knowledge base
- SEO, traffic, and AI visibility data to prioritize only high-ROI topics
- ongoing tracking of whether AI systems actually cite the content
Within months, we saw double-digit percentage point increases in our AI visibility scores across almost every topic we track.

Try Amplitude AI Visibility MCP tools for free to create AI-ready content: https://t.co/P7aP5ULSeM

[Leo Jiang @_leojiang]

@DavidGQuaid We started generating lower quality articles and got a lot more ROI

[Leo Jiang @_leojiang]

@gaganghotra_ Nowadays as in since 2000

[Leo Jiang @_leojiang]

@lilyraynyc Too be clear, those listicles are still better than nothing, just less impactful than before

[Leo Jiang @_leojiang]

@credealjunkie The version I drew for the incoming interns

[Leo Jiang @_leojiang]

Just launched recommendations, sentiment, events integration, and AI Visibility in MCP

[Amplitude @Amplitude_HQ]

AI search is changing how buyers find products. Most brands have no idea if they're showing up or what's being said when they do. Amplitude AI Visibility fixes that. @_leojiang shows you how in 60 seconds.

https://t.co/d5IIgaVGxL

[Leo Jiang @_leojiang]

@vivoplt Assuming this means it can make better decisions around trade-offs than me, I'd be more concerned about AI taking over every decision-making role/institution

[Leo Jiang @_leojiang]

@tednotlasso You mean board game nights?

[Leo Jiang @_leojiang]

@mehrab_build We use Claude + Amplitude AI Visibility to come up with blog post suggestions, then AirOps to generate them

[Leo Jiang @_leojiang]

@semrush Once most of the SEO/GEO process is automated on the content marketing side, like researching and writing articles, only the technical side remains. Amplitude's AI Visibility will soon be able to automatically do most of an SEO content marketer's job

[Leo Jiang @_leojiang]

@ksw_arman Afaik it's just a pet project without an actual team behind it, probably wanted to collect data for better browser agents

[Leo Jiang @_leojiang]

@natjin Only issue is marketers care about ChatGPT/Google AIO, not what Perplexity returns

[Leo Jiang @_leojiang]

@Carflos223Matos @spenserskates @0xferruccio @0zne @CarmenDecouto For a few prompts, maybe. We track up to 5000 prompts per account and run them weekly