writing code & fixing analytics. building sensorcore so your ai agent can actually do the math. i don't look at charts anymore.
Joined January 2015
- Tweets 733
- Following 22
- Followers 49
- Likes 48
1 Photos and videos
May 31
#stablediffusion, #nano #banana, #wan, #kling, #seedance… все что на #replicate есть. Видосы генери, картинки, голос в аудио... все бесплатно, никаких регистраций, карт, токенов, оплат. Одна страничка - выбрал модель и пользуешься.
#sensorcore . #dev
2
Apr 10
21 tools. One MCP server. Zero dashboards.
I'm building SensorCore — an analytics platform where AI agents are the users, not humans. And the entire product is organized into 21 tools across 4 categories.
Here's the full map
1
8
Apr 10
Remote Config (2 tools) — Feature flags from the chat.
- get_remote_config — read current flags
- set_remote_config_flag — toggle features in real time
The beauty: your agent picks the right tools automatically.
1
5
Apr 10
Ask "why did conversion drop?" — it chains Discovery → Data Access → ML Analytics. You don't orchestrate anything.
One config line. 21 tools. Any MCP-compatible agent.
Which tool would you use first?
8
Shipping an A/B test used to mean: build two variants, add a feature flag SDK, deploy, wait two weeks, export data to a spreadsheet, run a t-test in Python, argue about statistical significance.
1
11
Results look good? "Set new_onboarding to true for everyone."
Something broke? "Delete the new_onboarding flag."
One chat command. Instant effect. No deploy needed.
The whole A/B cycle — setup, measurement, analysis, rollout — without leaving your IDE.
1
5
How do you currently run A/B tests? Does it take weeks or minutes?
3
New users tell you about acquisition. Active users tell you about retention.
In the SensorCore Project Hub, these two graphs sit side by side for a reason.
New Users = "Are people discovering the app?"
Active Users = "Are they coming back?"
1
4
When you see active users dropping, you don't stare at the chart and guess. You ask your AI agent: "Why did active users drop last week?" It queries the actual logs, runs ML analysis, and tells you.
Dashboard for signals. Agent for answers.
1
10