Boris

@boris.bsky.social

Founder @getcensus.com with the quixotic goal of having every app & every workflow in a company run on the right data.

Has anyone created a standard format for structured *input* into LLMs? I'm thinking of a parseable json or markdown file that includes sections like examples, variables, system prompt, user prompt.

What is the word for when you’re debugging a data issue, you find one of the causes but it *increases* the scale of the discrepancy? 😂

I know we all expect it but it’s wild to watch folks outside of tech transition from using the verb “Google” to the verb “ChatGPT” in regular conversation.

Fun fact, Apple has the tiniest lobbying team in Washington. A fraction of the size of companies like Amazon, Meta, and Google. Has anyone studied or written about this?

Our team has been playing Timeguessr at our weekly all hands and it's pretty fun. We had to retire Geoguessr because one of our teammates utterly dominated it. Any other games we can try? TimeGuessr #540 43,651/50,000 🌎🟩🟩🟨 📅🟩🟩⬛ 🌎🟩🟩🟨 📅🟩🟨⬛ 🌎⬛️⬛️⬛️ 📅🟩🟩🟨 🌎🟩🟩🟨 📅🟩🟩🟨 🌎🟩🟩🟩 📅🟩🟩⬛ timeguessr.com

TimeGuessr

TimeGuessr - the game that tests both your geography and history knowledge. Guess the right time and location of a historical photo.

timeguessr.com

Would anyone use Google Slides if the collaboration features were removed? It's no surprise that folks like Canva & Figma feel like competing here – all the best presentation tools are single-player. Unbiased side plug: I think IA Presenter has a neat take.

Bild

Is anyone using Snowflake Cortex? Today's announcement makes it more compelling but my back of the envelope math is that it's between 4 to 20x the cost of using a generic Llama provider. Feels like a high cost for keeping compute inside their walls no?

As everyone feels the emotional rush of building a following here, useful to revisit Alex Zhu's brilliant "New World" analogy to explain the rise of TikTok. The key is social mobility. In established platforms like Twitter & Insta (ie. Europe), the social class system is already stabilized.

We've been automating a lot of RevOps via LLM enrichments on our data (e.g. sentiment analysis, fit scoring for sales, account research for outbount, etc). We've shared these prompts in docs/blog but feels like it should live in an open repo with contributions? www.getcensus.com/blog/sentime...

Customer Sentiment Automation For Outbounding and Support

Discover how AI-powered message analysis transforms customer sentiment insights for support and sales teams, turning nuanced data into actionable strategies across channels.

getcensus.com

After years of using Raycast, I still feel like I'm at the low end of "intermediate" user. What's everyone's favorite Raycast extensions, shortcuts, quicklinks, workflows, etc?

Agree it’s essential to embrace. But what is it about the data community that makes it resistant? Developers are a lot more comfortable embracing this. One theory? Generally weak testing / review practices so it’s scarier to trust randomness.

Jeremiah Lowin@jlowin.dev · 2y ago

I’m old enough to remember when data twitter got really really angry at the idea that AI would intersect their work. Maybe #databs is ready for real talk? Anthony Goldbloom nails it here with evidence, not hype: blog.sumble.com/the-overlook...