aayush4vedi

You are right that the adapter doesn't upgrade your old vectors — it aligns the two spaces, so retrieval stays bounded by what the old model encoded. If the goal is better representations on existing data, re-embed(btw my solutioning offers this dual-write path in migrateion as well :)

The adapter is for the more common case: the upgrade is forced (ada-002 gets deprecated) or a reindex is too big to eat at once — a billion vectors is ~$6k and a stale index for most of a day. Instead you cut over instantly: the new model serves new writes and queries through the rotation, the old index keeps serving untouched, and you re-embed in the background at your own pace. It's a zero-downtime transition, not a free quality bump. Small index? Just re-embed.

Changing your embedding model usually doesn't need a reindex, and that surprised me enough to build around it. Two models' vector spaces are close to a rotation of each other, so you can fit one orthogonal matrix (an SVD over a few thousand paired vectors, ~15s on CPU) and apply it to your query vectors at search time. An orthogonal transform preserves cosine exactly, so nearest-neighbour order is unchanged and the old index never moves. The whole "upgrade" is a 9 MB .npy.

Drift wraps that plus the unglamorous parts of running embeddings in prod: embed() dedups across runs and tracks per-run cost, watch() re-embeds only the rows that changed (Delta CDF, not the whole table), migrate() fits the rotation and won't ship it unless a recall check passes, and a SQLite ledger records cost and provenance.

Caveats, since you'll find them: today you apply the rotation to queries yourself (the drop-in client is next). It doesn't work for every model pair — GloVe→MPNet only hits ~71.5% recall, so migrate() measures recall and refuses below 0.97, falling back to a full dual-write reindex. embed() currently collects to the driver (fine to ~2M rows), pgvector is write-only so far, and the dollar figures are back-of-envelope.

Method's from the Drift-Adapter paper (EMNLP 2025, arXiv:2509.23471); I added the implementation and the recall gate. MIT, pip install drift-spark.

If you already run a nightly embedding job — what would actually stop you from replacing it? And which source/sink would you need (Iceberg, Postgres, LanceDB)?

Yes, My previous comment should make it a little more reasonable.

Eureka + Ikea = Eurekea

But it still sucks. I guess I'll be forefieting this domain & buying one with better name for the brand out of my pocket.

Thanks for the feedback @XCSme

Yes you are right-ideas are easiest to find. But finding ideas for which there is a demand & people are actually willing to pay for- is quiet difficult & iterative process.

My intention to build the product was to help people filter out already viable & successful products in their niche.And then just beat them in terms of better performance, pricing or by focusing on a particular feature or a user-segment. This will give the makers an unfair advantage to beat the competition out there.

"Good artists copy, great artists steal" - a quote Jobs stole from Picasso.

I agree that the placing & packaging of my product doesn't make it clear.Coming from Back-end Engineer background, this is quiet a challenge for me. Working on it for a re-launch :)

24yo, $5.5k Full Time Developer job, starting into the SideHustler world. Commenting here for my future (millionaire) self :3

@HN community, please advise me what am I missing with my side-hustle here.It's my first finished product & my main intention for launching it is to learn about the journey of a multi-hat wearing IndieHacker.

Hi Cory, this is really inspiring for the starting indie-hackers like me.Could you please share how do you continue building audience around your projects?

True.But these platforms are made to list the Software solutions which are trying to solve some problem-be it startup,saas,app or side-hustle. And the data I'm selling is for people who are trying to do the same. I've added demo spreadsheet on the site as well, you can have a look how the actual data will look like. Thanks

I am including all the latest items published on these two platforms, as of now: 1. ProductHunt(nearly 70K items) 2. Google Play Store(180K+ apps)

Though I have plans for including other platforms too & will decide after seeing response of this MVP launch.

Supp friend :)

I kind of build it as MVP-freemium & had future plans for additional paid features.But after such poor initial response,I lost hope & dropped it.Now that I'm less emotionally attached with it(after nearly 6 months & a little mature), came to realization that should've spent more time & efforts in marketing.Back then, I had very limited reach in online community(0 followers, near 0 karma etc).I am working on it currently and trying to make it a fore of habit.

The maker world is lonely indeed.Would love to get in touch with you.My contacts are available on my portfolio.Lets catch up sometime soon!

Riding in the same boat bro!

I tried to build & ship a simple tool [1] in one month(not 1 day btw), which got almost 0 customer engagement - just few comments on how good the idea is.

This made me question myself on solving a bigger problem & I spent around 5 months on a ML based product.Which eventually turned out out of my ML knowledge's reach (couldn't hire one too); so I stopped working on it & put it aside for sometime later.

Right now, like you, I'm working on building an audience on reddit/IH/twitter; so that I can ship my short-sprinted projects.

[1]: https://suppme.netlify.app/

Thanks! will check it out.

Amazing articles @PaulHoule! I'll dig deeper into them over the weekend. Are you planning to build some product using it though?Mayebe we can get in touch.

Hello HN!

Being a heavy tech-news consumer means you're reading this post in your n-th daily visit to reddit while taking a quick break from work.

Hundreds of dev-based news sources generate thousands of articles everyday; hundreds of brilliant bloggers & twitterites publish their content, along with the seldom unmissable discussion on the amazing communities.And just like any other tech-fanboy, I am in love with them.Its difficult to close that tab & go back to the work I was doing before :)

Although this 'hobby' is said to be among the most crucial things for growing professionally; there is too much of news & too little time.And most of it is spent in filtering out the relevant content & saving it for later(and sadly highly unlikely)read.

There exist many aggregators, newsletters and RSS feeds to clear this chaos, but there is generally a trade-off of either being too narrow for a certain tech OR having too much noise. Reviews of many such tools are filled with other issues like insufficient domain-filtering,lack of personalized content & being distracting-eventually missing the purpose.

My question for you is if x were the ideal tool which made your tech news reading life easier,what features would you like to see in x ?

  - Do you feel that the curated newsletters are a good for saving time OR despite having subscribed, you always found yourself lurking on news sites like Hackenews, Hackenoon, subreddits etc ? 
  - How do you keep tabs on the latest trends in your domain?
  - How much of cynicism do you want in x? Should you just be served the meaningful content for your domains of interest OR you want to explore all the content to find something new & interesting once for a while?Is it really worth the time & effort?
  - How do you feel about being charged for reading the otherwise freely available content?if yes, how much would you consider a moderate amount?
  - Please give your suggestions & pros/cons for the tools which are you using for this currently, if any?

The HN community is brimming with wiser, experienced folks and being a novice, I'd love any kind of feedback form you. Cheers!

Hello HN!

Being a heavy tech-news consumer means you're reading this post in your n-th daily visit to reddit while taking a quick break from work.

Hundreds of dev-based news sources generate thousands of articles everyday; hundreds of brilliant bloggers & twitterites publish their content, along with the seldom unmissable discussion on the amazing communities.And just like any other tech-fanboy, I am in love with them.Its difficult to close that tab & go back to the work I was doing before :)

Although this 'hobby' is said to be among the most crucial things for growing professionally; there is too much of news & too little time.And most of it is spent in filtering out the relevant content & saving it for later(and sadly highly unlikely)read.

There exist many aggregators, newsletters and RSS feeds to clear this chaos, but there is generally a trade-off of either being too narrow for a certain tech OR having too much noise. Reviews of many such tools are filled with other issues like insufficient domain-filtering,lack of personalized content & being distracting-eventually missing the purpose.

My question for you is if x were the ideal tool which made your tech news reading life easier,what features would you like to see in x ?

  - Do you feel that the curated newsletters are a good for saving time OR despite having subscribed, you always found yourself lurking on news sites like Hackenews, Hackenoon, subreddits etc ? 
  - How do you keep tabs on the latest trends in your domain?
  - How much of cynicism do you want in x? Should you just be served the meaningful content for your domains of interest OR you want to explore all the content to find something new & interesting once for a while?Is it really worth the time & effort?
  - How do you feel about being charged for reading the otherwise freely available content?if yes, how much would you consider a moderate amount?
  - Please give your suggestions & pros/cons for the tools which are you using for this currently, if any?

The HN community is brimming with wiser, experienced folks and being a novice, I'd love any kind of feedback form you. Cheers!

Hello HackerNews!

Supp is a simple browser extension for mindful status management. You can form group with your coworkers or friends, and Supp gives you an instant glance at what they are doing & who's free or busy.

The Idea

It’s designed mainly for the current remote/WFH heavy workspace, where working peacefully has, startlingly, become quite challenging.There are too many sync-ups, just to know what everyone is doing in the team.Along with frequent Slack/Hangouts messages like "Are you available for a call?", "What are you doing?". Not to mention some people who don't even bother to ask your availability before shooting out their pings.

I was unable to find disturbance-free time to focus on my own work and wished there was an automated way to know what everyone was doing & see if someone is free to be bothered with or not before pinging them.

Been there, felt that? Supp is for you!