Want to find a problem worth building a product for? Talk to more users.
I built Painfinder to let you chat with 25k user software reviews and find problems that nobody is solving. Powered by
@LangChainAI
from
@hwchase17
and
@pinecone
!
Here's a free idea hiding in the data ⬇️
Know where users get confused in your product?
No. You are an expert at using it
We built FastAB to show you where users get tripped up:
1⃣Upload a figma screenshot
2⃣Watch AI use your product 100s of times
Built
@southpkcommons
with
@replicatehq
and
@hwchase17
's
@LangChainAI
It’s hard to tell if NYT and Fox News even agree on reality sometimes
So we built Janus to show you where NYT and Fox find common ground, and where they don't. Created with
@TheMountainDu
!
Powered by
@LangChainAI
from
@hwchase17
.
See what they say about UFO 🛸 legislation👇
Very excited to share Eggnog with the world!
Every day coding and making whacky Eggnog videos with
@jitsvm
has been a ton of fun
Please go try Eggnog and share your videos with us!
🥚
@eggnog_ai
(YC W24) is building the YouTube for AI-generated content. Anyone can create, watch, and remix AI videos on Eggnog and build their own cinematic universe.
Congrats
@SamuelMGPlank
and
@jitsvm
on the launch!
In mad scientist mode rn.
I'm using FastAB to run experiments on simulated users.
We built it so that you can A/B test your product variants in 30 mins, not 2 weeks.
Powered by
@replicatehq
and
@LangChainAI
from
@hwchase17
.
See how 1k agents interact with my Reddit redesign👇
Stellar support from
@replicate
We had an influx of new users yesterday and hit the limit on number of models created, which caused our product to break
It took <45 mins from me sending a help message to
@fofrAI
to their fix hitting production and unblocking us. On a Sunday!
My gf has preemptively banned me from using her character to recreate any historical scenes from the roman empire.
She didn't say anything about the bronze age collapse though..
Thousands of searches have been made on Painfinder since I dropped it last week. Builders are hungry for real user problems to hack on.
Some of the most unexpected searches I've seen:
Added 350k more reviews to Painfinder. It's really showing in the results.
There are a ton more user pain points for builders to discover in the data ⬇️
@Integrity_Inst
is featured in the NYT good tech awards! We advocate for design changes, measurement, and transparency to make social platforms better for society. Kudos to
@sayhar
and
@jeff4llen
for creating the Integrity Institute and for welcoming me in
I'm presenting tomorrow night at the
@mlopscommunity
event in New York. Will talk about about simulating user interactions with products.
The other presenters look 🔥, definitely come!
Ask about 3D printing software, and Painfinder tells you that users have problems with STL files and 2D drawings. It will cite its work too.
Smells like opportunity...
Phoebe escaped out the back door yesterday. Then she turned around and booked it back inside when I went out to look for her
Now she's having second thoughts
The Air Head video from the sora blog post is so good because it gets you invested in the character. It's not just vibes.
And the filmmakers used a clever trick to make the character look coherent throughout the video
New age of consumer entertainment is going to be 💯
I was able to go from text -> full blown intergalactic alien saga video in under 10 mins using the recently launched
@eggnog_ai
Whether we like it or not, AI is here
On today's episode, I met
@SamuelMGPlank
, the creator of an
#AI
tool for podcasters. We talk about the impact of AI on everyday life, it's use in the legal profession and the future of
#ArtificialIntelligence
@jitsvm
This is what I used for the prompt: "chiseled jaw, piercing gaze, flawlessly ruffled hair (btw he got the highest score in distributed systems grad class)"
Do any of my followers make a podcast or have podcaster friends? I'm building a product for podcasters and would love to talk. I'll make you some fun AI-generated content about your podcast in exchange 🙂🎙️
Really not a bad outcome at all for a few minutes tooling about on - they make it a breeze to generate a few 4-second clips and then glue them together.
Backing track by Suno of course!
@EthanZ
I like the idea of being able to opt out of the ads-model by paying instead. But what about the risk that this would create a two-tiered web where affluent people can surf without fear of their data being exploited and everyone else has to live with privacy breaches?
We are thrilled to announce the launch of Red Line, our new internal platform housing long-form, investigative journalism. Fellows began work on these projects in February, and they will be rolled out over the next few weeks. Let us know what you think!
My thesis is about bot accounts that are integral to the anti-Trump resistance. Can't prove they are Russian, but it makes sense that they would play both sides.
@Karmedge
@LangChainAI
@hwchase17
@pinecone
Hey Robert -- I pulled a bunch of the most similar reviews from the vector DB, then did a first pass over those reviews with GPT 3.5 to filter down to only text that was relevant to the query. Then I put only the relevant text into a GPT-4 call. A bit expensive, but manageable.
@paulg
We think AI will also make it possible to run experiments on simulated users that act like your users. The product development loop will get even faster.
In mad scientist mode rn.
I'm using FastAB to run experiments on simulated users.
We built it so that you can A/B test your product variants in 30 mins, not 2 weeks.
Powered by
@replicatehq
and
@LangChainAI
from
@hwchase17
.
See how 1k agents interact with my Reddit redesign👇
Janus pulls in articles from both NYT and Fox that talk about the same story.
It summarizes the shared context at the top so you can see what they agree on.
And a rogue one for kicks:
"Can we develop a mind-reading software to predict and track spontaneous outbreaks of impromptu musical numbers in public spaces?"
The simulated users in the test group clicked upvote 36% more than the control group.
The redesign must be working!
The results tell you how every metric moved and put confidence intervals around each point estimate.