HyperArc

A Digital Clone Of Your BI Team

[HyperArc @hyperarc_inc]

HyperArc is now part of Zendesk! Together, we’ll accelerate innovation, enhance AI-driven customer experiences, and expand analytics expertise to deliver exceptional service for every person on the planet. Follow @Zendesk on all socials for the latest.

[Zendesk @Zendesk]

Big news: Zendesk has acquired HyperArc, an AI-native analytics platform known for its HyperGraph engine and real-time GenAI insights: https://t.co/Ed2QCrkWPT

This is a major leap forward in our analytics roadmap — bringing faster, smarter, and more powerful insights.

[Zuye Zheng @zuyezheng]

OpenAI's deep research allowed users to beyond single questions and into complex multi-step problems, giving it the opportunity to understand what's still missing and retrieving it between each step.

@hyperarc_inc's Deep Exploration and Narrate brings this to analytics with human-in-the-loop planning, multi-query execution, self-critique with replanning, and story telling across queries. Find out more in our post!

https://t.co/4JSNmiay3k

[Zuye Zheng @zuyezheng]

Check out the future of data with analytics built natively for AI with @hyperarc_inc's launch on Product Hunt.

https://t.co/hQY7p2Iak3

See the magic with questions on our public datasets at https://t.co/vHEU3McZs9 before you try it out for free on your own data.

[HyperArc @hyperarc_inc]

Try out our MCP server with your analytics.

https://t.co/Xfof5bmnIW

[Zuye Zheng @zuyezheng]

Agents + MCP is the Future of Work

Slack replaced email as “the future of work” with its integrations and bots — transforming pure messaging to something that was actionable.

Agents such as Anthropic's Claude Desktop supporting MCP is to Slack what Slack was to email.

See how @hyperarc_inc's MCP server can empower your agents to make the same data driven decisions as your humans.

https://t.co/TmZZcRfQTQ

[Zuye Zheng @zuyezheng]

How @hyperarc_inc taught LLMs math to be best analyst.

https://t.co/zoIDUx66lH

[HyperArc @hyperarc_inc]

We're excited to launch HyperArc themes with our 2 newest customers, Lumon (who will be helping define our memory sharing road map) and Catbus Inc. (a referral from our neighbor Totoro).

[Zuye Zheng @zuyezheng]

Now that you've seen the magic of Vibe Coding from @karpathy, check out Vibe Analysis with @hyperarc_inc.

https://t.co/YV1dM9tQtf

[Zuye Zheng @zuyezheng]

Trending US Executive Orders with LLM Data Pipelines

Using LLMs to quantify and cluster all Executive Orders since 1791 for sentiment and political bias.

We ran an experiment using embeddings and a zero shot, prompt based technique for sentiment analysis including analyzing each text along 7 political axes: Ideological Tilt, Immigration, Executive Power, Federalism, National Security, Economic Regulation, and Social Inclusion.

https://t.co/liAkKFyZze

[Zuye Zheng @zuyezheng]

What if your analytics had web search like ChatGPT? What would you ask it?

https://t.co/NTA8HpHsUZ

[HyperArc @hyperarc_inc]

RT @zuyezheng: Everyone's been talking about the up to $500M in US State Department procurement to @elonmusk for armored Tesla's, but whose…

[HyperArc @hyperarc_inc]

See how HyperArc is building the Cursor of BI by reimaging how users interact with BI in a world of AI.

[Zuye Zheng @zuyezheng]

What does it take to build the Cursor of BI? We think it needs an entirely new UX paradigm instead of bolting chat onto traditional BI like Tableau or PowerBI.

https://t.co/OSaCpd3Cv3

[HyperArc @hyperarc_inc]

Find out how to empower your entire team with grounded and personalized natural language queries and try it out free at https://t.co/eD2v1mxKqr.

[Zuye Zheng @zuyezheng]

On a recent benchmark, frontier models like o1 only achieve 17% accuracy in text to sql. A brief history of natural language queries and a new approach using memories and prompt expansion like planning with @hyperarc_inc's Continue Exploration.

https://t.co/1PRVfbz0i1

[HyperArc @hyperarc_inc]

A new way to search in enterprise #BusinessIntelligence.

[Zuye Zheng @zuyezheng]

How's everyone's search experience in #Tableau, #PowerBI , or even @googledrive?

One of the biggest problems we had at Salesforce was a proliferation of duplicate reports (and work needed to create them) since you could never find the one you or your team had just created. Did I call my sales report "New Opps" or "Pipeline"?

Memories in @hyperarc_inc give us rich, automated, natural language descriptions of each query that goes beyond the title and into the intuition and insights found. Our natural language search across all our assets now leverages memories so search for "gobble gobble" or "territories with the most new deals" and find exactly what you're looking for!

[HyperArc @hyperarc_inc]

How would you incorporate forked memories in your BI workflow?

[Zuye Zheng @zuyezheng]

https://t.co/9PpWn2JVug

[Zuye Zheng @zuyezheng]

Notion + Tableau

What happens when you get to precisely annotate and augment your data with knowledge directly in your BI tool? Your search and natural language queries gain superpowers.

https://t.co/TG89UFQMHD

[Zuye Zheng @zuyezheng]

Handling unbalanced datasets when using AI in analytics.

https://t.co/fYL5EFpAPf

[HyperArc @hyperarc_inc]

Inspire creativity with the HyperArc Agent.

[Zuye Zheng @zuyezheng]

The benchmark for AI vs human is often creativity with current LLMs sometimes compared to a stochastic parrot (which lets me sleep a bit more soundly at night). This lack of creativity is also often associated with a lack of utility, but maybe we're thinking about creativity with AI wrong?

What we've found internally using tools like @ChatGPTapp and @cursor_ai chat is they do lack creativity, but their super power is actually to inspire creativity in humans. Their ability to bounce ideas off of and suggest that ember that becomes the aha moment to push your work to the next level is what gets us to come back.

This was the inspiration for the @hyperarc_inc Agent, a digital copy of your BI team that's always looking over your shoulder (in a non-creepy way) and ready to provide that ember of creativity, grounded by the memories of all past explorations. Understanding the semantics of age, we're able to suggest correlations with higher education that will push your analysis further.

While building this, we also found a pretty interesting side-effect, the agent started learning product "hacks" from user interactions in memories in addition to just next best analysis. With only local geospatial features, @tao_tristan and I started rounding longitude/latitude coordinates for an approximation of clustering, which to our surprise made it to suggestions.

[Zuye Zheng @zuyezheng]

Since @github is down... thought I'd share something interesting we found integrating AI with BI at @hyperarc_inc .

We haven't had much success in using AI to "solve" BI, but we've found AI to be very helpful in explaining the intent behind BI in a very human way when augmenting an analyst's exploration.

With an Olympics dataset we are able to use our HyperGraph to automatically generate the "human" intuition behind specific (key is specific) changes such as filtering between specific years (1939-1945) to examine "Participation During World War II". Or for specific countries (USA, Mexico, and Canada) to examine "North American Olympic Sports Participation".

With this intuition annotating tabular data with natural language we're able to power a deeper semantic search grounded on past explorations.

See some deeper AI assisted exploration from Tristan Tao using HyperArc: https://t.co/pFQ7RNt7lM

[Zuye Zheng @zuyezheng]

A metrics layer is like your New Year's resolution…

It's a great idea, you should probably do it. But like your New Year's resolution, it’s two weeks into January and you've probably already started taking out some of that ice from your daily cold plunge, those intermittent fasts are becoming a little more intermittent, and you’ve been staring into the sun a little less every morning while listening to Huberman.

Unlike your New Year’s resolution you can automate your metrics layer with HyperArc.

https://t.co/3K25Pr5AcV