Hi HN, we built this because running SAM3 on long videos without a GPU was painful due to memory constraints.
The core idea is automatic overlapping chunking based on available RAM, with IoU-based object ID remapping so tracking remains consistent across chunks.
Great question. PI seems to count the 'number of clicks on your pin' and Reddit does the same. I have the same feeling that they overcount the number to make more money. On more occasions than one I've written to cust support asking for clarifications, the only one that comes through is that 'maybe they bounce as soon as they land on your page'. I find this strange, since the interaction time is 0 secs in almost every case. I've now been warned by several marketing wall of famers that Reddit ads is a scam :)
Unless I'm reading this wrong: is this a search engine that indexes everything Google doesn't?
I can see this being valuable for closed groups that define their own objectives i.e. a SubReddit-esque set-up where one can search for information on a theme say finance. The quality of results on the meta level risks to be lower.
On second thought, the idea of Google now indexing my personal info because I hooked onto the graph? Not sure how comfortable I am with that. Intriguing if Google does not index it.
Cool idea though, good luck with it!
Hi guys, I'm a self employed building an app called Penguin to help combat the problem of irregular pay amongst the self employed. We’re pretty nascent and struggling to understand if this would help the gig community. We’d be indebted for your take on it if possible, thanks a ton :)
I run a legaltech in the UK. Other than dealing with a similar case with my own co-founder (I ended up splitting ways very early which in hindsight I'm glad I did), we've dealt with a case on the platform where one co-founder wanted to oust the other for similar reasons. Not even thinking of this with a legal lens, if trust is already an issue, you need to have a 'hard conversation' with him. Trust is everything at an early stage, more than the problem.
Not sure I agree. I've seen many repeat founders showcase their product without mentioning their credentials. Here are the D3js founders showcasing ObservableHQ, and commenters discovering the team post a background check, no mention of their background anywhere:
https://news.ycombinator.com/item?id=16274686
On a side note, and no pun intended, I find it interesting that you find the need to mention that you are a ex-YC company to draw attention towards your offering. I'm curious to know how many upvotes this launch would have have gathered once you take that out. Just saying.
Jove is a financial concierge service for the self employed. We're trying to make financial products cheaper, transparent and accessible to freelancers. The free messenger beta launches with income protection insurance. It also features a COVID calculator to help freelancers access various government help schemes. We're very early into this, and would be extremely grateful for any feedback/suggestions!
Not sure where the impression of giving a cut is involved, there is no mention of that in the description?
Re your point, developing ones brand does not discount the possibility of falling into a blind spot anyway, it is one of the side effects of a sporadic cashflow lifestyle.
Hope you're all keeping safe in these trying times. We've been hearing stories about cashflow problems for freelancers and startups due to Covid. This is obviously something that causes a lot of added stress for them especially. We've been mulling over how we could help and decided to do the little we can of offering a free legal helpline to aid them in understanding how to proceed with the business interruption loan announced by the UK government.
By calling our helpline at (+44) 1274 79 2693 or messaging us, they can expect a response to their question instantly.
If you know any startups/SMBs at struggling with this, please feel free to share this helpline number with them!
I run a legaltech called 40Q, happy to help and figure out if there is something that can be done here. Is on a case to case basis. Feel free to reach out.
The reason I'm writing here is because I'm constantly surprised by the relative ease with which one 'lands' a 100k per year contract, or accidentally makes an API that forks its way to glory. I'm fortunate to know some amazing founders, some of whom are raising millions in VC. Been in and out of accelerators myself. Yet no story I know starts or has ever been like that.
Imho, it is excruciatingly hard work to provide people value and get them to pay for it. Takes time, constant follow-ups, a strong value proposition..none of which gets built jlt.
I wouldn't be fooled by the end result of $ x M ARR. There is a lot going on behind the scenes, which rarely gets spoken of because it sounds just sounds rad to say that I had an idea and somebody just signed me a cheque. No matter the entity at the paying end, people are fundamentally programmed to be uneasy letting go of large sums of money. Takes a lot of convincing to get there.
Having said that, building is an addictive hellride and I wouldn't trade it for anything else :)
Happy to share more of the limited experience I have, feel free to PM. Cheers to building!
The core idea is automatic overlapping chunking based on available RAM, with IoU-based object ID remapping so tracking remains consistent across chunks.
It supports:
- Text/point/box/mask prompts - Video + image inference - Segment-level processing - Built-in profiling
Would love feedback -- especially from people running segmentation workloads on constrained hardware.
Thanks! Prashant & Ankuj