Hari Demirev
2.4K posts
invented a new drink, the “low volume martini”, it’s basically any martini made in my apt
English
@james406 the windows start to feel more and more arbitrary the bigger they get, good luck getting good recall or accuracy if you dump in more than 100k tokens
English
it's wild to think about how massive 1M token context windows in LLMs really are
that's roughly equivalent to:
- the complete works of Shakespeare
- 11 hours of audio
- VCs telling you how they got into investing
English
@MatterFormula your payment processor has your MCC misclassified as convenience stores / speciality markets so I’m only getting 1x points!
English
three years ago I had a vision and today it was fulfilled @MatterFormula
Hari Demirev@hdemirev
I should be able to specify a ml quantity of dressing and sweetgreen should have some machine that dispenses exactly that amount, every time
English
$199/month?
I built 80% of this (local Miami only) in 30 days for $500 w AI and charge $19/month.
OneGuyMiami.com
Casa@getcasa
After two years of building under wraps, today we're announcing Casa – your personal property manager. We've raised $27M to redefine the homeownership experience from the ground up. We believe your home is your most treasured asset, emotionally & financially. It shouldn't also be a second job. Most homeowners are on their own – expected to have the time, expertise, and relationships to keep things running. Finding a plumber you can trust. Remembering when the HVAC was last serviced. Knowing what's actually wrong before someone shows up to fix it. Casa gives every homeowner what used to be reserved for the few: a dedicated team that knows your home deeply, handles the work, and stays in your corner. We're enabling this by building a deep, technical understanding of every home we serve – something that's never existed before, across 100 million single-family homes in the country. For $199/mo, membership includes: - A complete inventory of your home, built using specialized hardware & software - 1.5 hours of handyman time every month (and it rolls over) - Unlimited Concierge requests to take on virtually any home project - Custom, proactive care plans built specifically for your home - Weekly package and donation pickups - Scheduling and payments for your regular vendors - Utility and property tax monitoring …and we’re just getting started, with more benefits on the way to make the experience of owning your home as magical as it always should have been. Available now in the SF Bay Area and Los Angeles. Reserve your spot everywhere else. → getcasa.com
English
the limitation of liability era ie everyone is solving for the lowest common denominator bots and bad actors and you have to fill out a captcha to chat with the lululemon virtual assistant
English
@ron_ecomm just came across this, would love to chat more - DM'd you
English
We have been spending time working on our data labelling process of our CX department for exactly this.
Despite all the hype with AI, I have found very few actual areas to implement it beyond copywriting and some AI gen images/b-rolls.
One area we are focusing on is standardizing all the tags for all customer support tickets (emails + socials).
It actually requires a lot of time and effort to define these tags in a way that's relevant for EVERY single department.
You first need to understand the workflow and internal logic within each department and then from there define the tags accordingly.
The tags need to be at the exact granularity so that it will generate meaningful signal for other departments to action on.
It's as "cross-functional" as it can possibly be.
One example I have found this is ABSOLUTELY crucial for is product quality control.
Do you know how difficult it is to identify how your garment fits on your customers? It's nearly impossible unless you have massive amount of data. This is why all the pants you buy fit differently. The entire fashion world is just operating on "close enough".
Your customers have already told you why it didn't fit them during the return process but are you taking all that data to update your spec to optimize towards a lower return/exchange rate?
And this is just one little thing in one department.
The end state to work towards is the ability to generate every single signal you can possibly extract from every single customer interaction you have, backed by data, to inform of your future strategy in both supply (new products) and demand (new ads) while making sure you have absolute product quality control.
This is THE edge as an internet business that you have to sharpen in order to compete against the non-internet businesses IMO. This will make you more agile as a business because the entire org is reacting in real time to market feedback.
p.s. kinda hilarious that I don't think I have ever heard of anyone on DTC X talk about quality control LOL
Connor MacDonald@couuor
The quote that most resonated with me from this ep: "For most brands the actual capacity to make AI or any humans effective depends on the quality of that underlying infrastructure... [brands] don't spend enough time on the organization of the databasing of their information, as an organization, in order to enable fast movement and effective team function." At Ridge, the last few weeks have felt like an inflection point for AI automating and augmenting what were already tight processes and a clear view on the data that matters to our business. It's easy to get caught up in new shiny tools and AI apps, but what has been infinitely more impactful are the AI workflows that tap into and use our underlying data infrastructure. The Saras Analytics MCP, for instance, will likely be the most impactful AI-related tool we'll use this year. It takes years of thoughtfully formatted and structured data and makes it more accessible and actionable than ever. With a few prompts and well-documented skills, we'll automate hours of work and have a more granular understanding than we'd otherwise have. And it is no different than the ad naming conventions in Northbeam, or the Creative Key gSheet we use for tracking ads produced; It all suddenly feels super-charged The episode is cool because there are multiple very clear examples of how this is being done at CTC — I took plenty of notes for myself.
English
always twelve steps ahead but it’s ten steps in the wrong direction because I thought that at step two somebody would throw a glass
English
@ndrewpignanelli me when I ask cursor to check if a date is in the past 24 hours
English
@anuraggoel @render the email doesn't say, but it's going up by 12%. would be handy to be able to preview the diff or something w/ a past period. mb it's custom domains
English
BIG pricing update for @render. No more seat fees, and more self-serve security and compliance capabilities.
Render@render
We're removing seat fees from our pricing. Our new Pro plan is $25/mo, flat. Add your entire team at no additional cost. We've structured our new plans from the ground up to make pricing fairer and more scalable for teams shipping faster with AI. What we're changing 🧵
English
It's the president's war on snacks. Bonne Maman preserves is now $7.99/jar at Whole Foods in New York. Was $6.49 a year ago. A Pain D'Avignon baguette is now $4.79, up from $4.50.
Scott Lincicome@scottlincicome
"Trump ended free trade for Mexican tomatoes. Prices are up 23 percent in the last year." reason.com/2026/04/17/tru…
English
checking in here
just following up
quick ping on this
is this still top of mind
any eta you can share on this
hey just checking in
can you provide a status update
what’s the latest on this
wanted to make sure you saw this
any update on this
let me know how we can push this along
English
ok so I tried to find usable benchmarks, couldn't, made my own. turns out @cerebras is the fastest, @togethercompute the slowest / lots of 503s
English
a small child has barricaded the pret door from the outside. this is as close as I'm ever going to get to dakota johnson blue bottle rope sitch
English
@ZackKorman weird niche overlap where I was genuinely excited here for a sec bc I am in the market for embroidery digitization software…
English
