Zairah Mustahsan

ai & evals @

[Zairah Mustahsan @zairah___]

It was a pleasure to speak with @KashmirLife about my education journey and discuss the current state and future of Data and AI. Thank you for inviting me!
Full episode 👉 https://t.co/5pWrxaUdHM

[Zairah Mustahsan @zairah___]

@seyarkarimii welcome!

[Zairah Mustahsan @zairah___]

Our team at @youdotcom has been shipping!!! Check out our new feature, Search with Highlights, which makes your agent more accurate and less costly!

[You.com @youdotcom]

New: Web Search with Highlights. 95.17% on SimpleQA. Same $5 CPM. Independently verified by @braintrust: 35% cheaper than Claude's built-in search, 11% cheaper than OpenAI's.

[Izzy Hurley @iz_hurley_]

when I was in middle school we had, like, a whole month dedicated to learning how to google effectively... and tbh I think I got pretttty good

this eval kind of felt like the final quiz from that unit, except instead of students it was a bunch of web-search-agent configs

[Zairah Mustahsan @zairah___]

Get started docs: https://t.co/B7nL2jXnFO
Artificial Analysis detailed blog: https://t.co/AGZOa2ZIzC

[Zairah Mustahsan @zairah___]

If your agents' web search feels slow or inaccurate, it's time to upgrade: @youdotcom web search just topped @ArtificialAnlys benchmark across 14 search APIs with a 74 Search Index score in 18 seconds. No one else matches that speed-accuracy combo.

Find our get-started docs 👇

[You.com @youdotcom]

No web search API returns more accurate results faster than @youdotcom Web Search with Highlights.

Artificial Analysis benchmarked 14 search APIs across three evals. We score 74 on their Search Index in 18 seconds per task.

No other provider beats this accuracy without doubling your user’s wait.

https://t.co/u3QNa3OBRq

@ArtificialAnlys @mrdbourke

[Zairah Mustahsan @zairah___]

@mrdbourke @ArtificialAnlys @mrdbourke relevant doc https://t.co/bqa312pudA

[Zairah Mustahsan @zairah___]

@mrdbourke @ArtificialAnlys I own evals at https://t.co/3PARTKZLEa, curious to see where our new ‘highlights’ extraction mode lands here. Could you add that?

[Zairah Mustahsan @zairah___]

Instagram and meta have totally lost it. On IG showing me what a friend is selling on fb marketplace.

Not only is this invasion to privacy but also why do I need to know this on instagram?

[You.com @youdotcom]

In one week (8/13) We're hosting the @youdotcom AI Forum at @Databricks in SF full of people who build AI.

Join us if you're interested in talking through systems that break in production and the model lock-in nobody admits to.

Approval-required. Request to join 🛠️

[Zairah Mustahsan @zairah___]

Life in AI space: go to sleep with a todo list… wake up to a whole new world… replan… eat something… quietly lose your mind… repeat

[Zairah Mustahsan @zairah___]

Our team at @youdotcom just hit SOTA on DeepSearchQA: 83.67% accuracy and 93.16% F1.

💡The core insight: multi-agent setups consistently lose to a single well-engineered agent, given the same compute budget.

🛠️The fix? Treat the budget as a first-class citizen of the reasoning loop. When the agent knows its remaining tokens from turn zero, it plans differently.

📈We scaled to 10M tokens and 1,000 turns in a single session to get there.

🗒️Full breakdown of the harness architecture 👇
https://t.co/esXFfuVEyv

🤔Curious: Are you still defaulting to multi-agent for complex search tasks, or have you moved to single-agent scaling?

[You.com @youdotcom]

Multi-agent swarms underperform a single well-scaled agent at the same compute budget. 🧵

We built around that finding—and hit SOTA on DeepSearchQA: 83.67% accuracy, 10M tokens, 1,000 turns.

🔗 https://t.co/g0aPXdsHt1

[Zairah Mustahsan @zairah___]

@LLM_eval Hi
@LLM_eval
could we get an extension for full paper submissions? By EOD Sep 7th?

[Zairah Mustahsan @zairah___]

I've spent my free time in the last decade studying quantum mechanics, relativity, and neuroscience to explore whether our experience of time might be an illusion.

Something made me think this past weekend: Within minutes @youdotcom 's Advanced Research and Insights agent generated a thorough 15-page report that stitched together intricate concepts from literature and quantum physics that had taken me years to grasp.

https://t.co/cmKV0piP7l

Compared to the slow, deliberate pace of human cognition, the speed at which AI is capable of advanced research is fascinating. This makes me think of Virginia Woolf's "Mrs. Dalloway" in which the characters' perceptions of time vary according to their mental states and life experiences; some moments seem to go on forever, while others seem to shorten.

The report notes that "our subjective experience of time, including the sense of a flowing present, may be a useful illusion created by our brains to navigate the world"—yet here's an intelligence that processes information in ways unbound by our neural constraints. This presents an intriguing paradox: will our perception of time itself change as AI collapses the time needed to synthesize complex knowledge?

As quantum experiments show us that "at the most fundamental level of reality, time may not flow in a single direction," the rapid learning of AI can help us rethink how we now perceive time.

It's fascinating to me that while studying the idea that time is an illusion, I've come across an AI technology that alters my perception of time, making me question not just what I know, but how time shapes the very way I come to know it.

#ai #technology #meaningoftime

[Richard Socher @RichardSocher]

Wow. This is huge news for YOU.

https://t.co/R3fWmYJZXa will be a default browser option for EU countries on iOS 17.4

We're the only chat-native browser option that made the cut.

Read more: https://t.co/pUwMMM6hEU

[Richard Socher @RichardSocher]

Want your chatbot to be up-to-date?

Our new YOU API empowers your LLM with real-time web search, much longer snippets and better fact recall than any other search API.

We also have pure news endpoints and complete retrieval-augmented generation (RAG).

more details and links >>

[Zairah Mustahsan @zairah___]

If you are #odsceast today come stop by Ballroom A at 3pm to learn about best practices in #dataengineering and tips to set up a Lakehouse Architecture to drive your organization's data success! #datadriven #DataScience
@_odsc

[Zairah Mustahsan @zairah___]

Try 👉 https://t.co/a22htWdjWM

to get started!

[Zairah Mustahsan @zairah___]

So grateful for the amazing co-workers who are not only building the next gen #GenerativeAI products but also send kind messages using that technology 🤩

Try https://t.co/mFM5gn71O4 & comment below your favorite generations to express #gratitude to your female co-workers💌