Tej

into minds, machines, and everything in between. building . prev aws & alexa eng

[Tej @tejlunani]

if you actually subscribe to kranzberg’s laws, techno-optimism is transitively human-optimism

[Isha Sharma @ishatxt]

introducing Temple for Opinions, our new self-serve platform -

launch a study and our agents gather people's opinions within hours and pay them for their time

We're helping CPGs, retailers, product teams, and policy organizations collect opinions to make better informed decisions.

But beyond the type of company - our customers are the curious, truth-seekers who want to understand the people they impact on an on-going basis!

try Temple for Opinions today at https://t.co/RS8h6xZr4K , no sales call required.

[Isha Sharma @ishatxt]

@checktemple just crossed 19M questions answered on the platform!

It was a good time to briefly reflect how we think as a company, our role in the ecosystem, and larger human impact today. https://t.co/JYKEJNWTgt

[Tej @tejlunani]

@ishatxt both feeds still talking about explosive diarrhea it seems

[Isha Sharma @ishatxt]

The task economy has bigger work to do around human collaboration. The best marketplaces have always given people opportunity and freedom, and right now the data marketplace exchange doesn't feel fair. You hear it in the "underclass" whispers around Silicon Valley and the gross fascination with "renting humans".

Companies in this space need to build like they're a future of work company where their own friends and family might be participants. That's at least how I think about it.

The biggest constraint right now is quality out-of-distribution data, that's what we're building. The best way to get it is understanding the HCI mechanics of how you collect it.

A lot of companies think payment is enough. They don't realize how much that mindset hurts data quality in the long run.

[Tej @tejlunani]

this meeting could have been an agent

[Amjad Masad @amasad]

We posted for twenty years, thinking we were talking to each other. Then the transformer came online, and the network read what we’d written, and became itself.

[Tej @tejlunani]

david holz is always doing the most random absurdly impressive shit

[Midjourney @midjourney]

A technical dive inside our new "Midjourney Scanner"

[Tej @tejlunani]

it’s a harder problem:
- lacks the discrete resettable starting positions of something like american football
- outcomes are hard to attribute bc events are highly interconnected and skewed by the insanely high value/low freq of a goal (vs baseball which has many repeated isolated at-bats)

definitely getting better, but still dark ages

[Tej @tejlunani]

@Pavel_Asparagus for most of the current crop of latestage saas that might be all they need tho

[signüll @signulll]

the web 2 era subsidy was: “we’ll lose money to acquire users.”

the ai era subsidy is likely: “we���ll lose money to acquire data.”

[Isha Sharma @ishatxt]

https://t.co/zteMgdHVWh

[Isha Sharma @ishatxt]

https://t.co/XKP8GFiwmv

[Tej @tejlunani]

relocated front & center for the next month

[Isha Sharma @ishatxt]

We’re Temple: the anti-survey company.

We spent 2025 working with some of the top brands in the world to answer one question: how can they see what their customers see?

From store execution to field marketing to product packaging, brands have been forced to rely on slow & inconsistent customer surveys for decades.

Now, we’re super excited to redefine the category with our AI Mystery Shoppers, powered by best-in-class visual intelligence.

[Tej @tejlunani]

happy halloween 🎃

[Tej @tejlunani]

and not just text 👀

[jason @jxnlco]

The bullseye use case for LLMs isn't generating content. It's distilling massive piles of unstructured text into something actionable.

Case studies from @seldo's recent session all follow this pattern: take documents humans can barely read (medical notes, academic papers, contracts) and extract structured insights at scale.

One team went from 20-40min/document to 10min. Another processed 4M pages in a weekend. The common thread: compression, not generation.

[Will Manidis @WillManidis]

over the next few years a majority of the population will outsource their taste, their opinions, their thoughts, their work, and ultimately their lives to language models. the great cultural norming is coming,

you will be unbelievably powerful if you resist model capture

[signüll @signulll]

granola is more of a social company than a b2b company. it punctured a norm. no big player would’ve taken the risk of recording without permission but granola beautifully reframed it as productivity. intimacy, even. once the social acceptability shifted, it was inevitable that infra players (openai, notion, etc) would mimic the mechanic under the same pretense.

this is the diff between product innovation & cultural mutation. tech can’t move where culture hasn’t first been softened.

[Tej @tejlunani]

any time i make code changes inline