282 karma · joined November 9, 2022
Caveat here is you may not be able to use their environments or you may or may not have their kind or robots to train your roomba. But at-least you could get an idea of how RL training is done for robots like yours.
0. Emails suffer from a "misclassification" of intent issue on a time*attention scale. Imagine time of the day/week/year on one axis and their attention on email inbox on the other. Emails have to arrive at the right (x,y) point for a user to act on. But they rarely do.
1. Well being of a user is proportional to their current state of mind to receive an message from X. Which is proportional to how likely they are to listen what you have to say.
Both of these suggest a negotiation of messages between two parties, much like when a bartender asks you if you want a refill and you can say yes/no.
We could use two negotiating agent, e.g. my agent that knows what I care about now/today/1-week ago and negotiates with an aspirant sender's agent before they send me any messages. e.g. I could set a policy based (my ToM) for my agent like "Between 1-1:15PM every day I want to read about all product announcements I subscribed to for XYZ product type". My agent would go talk to the aspirant's sender agent and gets messages right then.
An alternative policy could be "I have some free time now, create a summary/gist of all announcements on products I might be interested in.". The agents would negotiate with the sender to do the same.
Signups emails would be to replaced by an agent which "creates" a ToM with sender on hard-stop dates. I would tell my agent : "I am interested in this logging service to compare different ones, I will not be interested once ENG-123 is closed" and mine would not just tell the sender that they are not interested when the time comes (which is when ENG-123 is closed).
Longer term policies would just age out any message negotiations because I don't like/care about those products anymore.
For chatgpt desktop and Claude desktop my experience with MCPs connected to my home NAS is pretty poor. It(as in the app) often times out fetching data(even though there is no latency for serving the request in the logs), often the existing connection gets invalidated between 2 chat turns and chat gpt just moves on answering without the file in hand.
I am not using it for writing code, its mostly read only access to Fs. Has anyone surmounted these problems for this access patterns and written about how to build mcps to be reliable?
Here is how I built it in a WUI: I sent SSE events from Server -> Client streaming web-search progress, but then the client could update a `x` box on "parent" widget using the `id` from a SSE event using a simple REST call. The `id` could belong to parent web-search or to certain URLs which are being fetched. And then whatever is yielding your SSE lines would check the db would cancel the send(assuming it had not sent all the words already).
Instant crosses that persistence boundary, your app can propagate updates to any one who has subscribed to the abstract datastore — which is on a server somewhere, so you the engineer don't have to write that code. Right?
But how is this different/better than things like, i wanna say, vercel/nextjs or the like that host similar infra?
Its quite hard to believe why it took this much inference power ($20K i believe) to find the TCP and H264 class of exploits. I feel like its just the training data/harness based traces for security that might be the innovation here, not the model.
I would begin in one room to practice this.
All of these technically could have zero day vulnerabilities and people/companies who made it don't have the resources to buy 20000$ of tokens to go debug them... Maybe they don't care but if they do, what if they can't afford such models or get access in time.
I would like to know how can someone like me defend against them?
Its quite surreal. Very much in-your-face concrete exposure. Yet, to walk and experience it with your eyes is a study of contrasts: a giant, comparitively modern, greenhouse, has a glass roof open to the sky and yet many floors have no light or windows at all. And in the outdoor spaces, like the fountain/canal running through the complex the concrete will sort of be in the background and lets you focus on everything else: the water, the swans and the people around.
Juxtapose that to low hanging exposed concrete roofs and walls in closed passages could make one feel constrained/claustrophobic/yearning for light.
Part of the reason handy is awesome is because it uses some of the same rust infra for integrating with the model, so that actually makes it possible to use the code as a library in android or iOS. I have an android app that runs on a local model on the phone too using this.
Like with 10ms then online replication/backup — analogus to litestream for sqlite — but for in memory processes becomes feasible, no?
I'll say this is still quite useful win for browser control usecases and also for debugging their crashes.
A recent nationwide poll[1] shows AI has a poorer approval rating than ICE — ICE! — probably due to their overlords being "those" SV types. Everyday AI features are being shoved down our throats. I can't even choose to not install Gemini related apps on my Android when I select "which apps to install" when booting a new phone.
But people are a weird bunch. They largely don't buy products aligning with their values. No one is jumping up and down for Graphene phones even if they had amazing privacy first software. People buy 6mi/gal hummers and iPhones for fashion, brand, money, convenience/function. The pain threshold of all bad effects still is not high enough to quit their products in a meaningful way. Values and privacy are way down in their list. I wish people would not buy/install AI related features by big tech and be more discerning, but that is likely a pipe dream.
[1] https://pos.org/wp-content/uploads/2026/03/260072-NBC-March-...
I would say this is a values mismatch, if the client does not care about prioritizing technical debt(no matter how it got created) and that is leading to too much mental angst, I would recommend leaving the project. But before you make that decision, think about what you said
"They still want me to participate in the project and work on the most critical parts of the application"
I think in this new world, the job of software engineering is to identify where AI could be used to provide the highest levers to the most important people in the client's team to make decisions. For example, can AI be used to identify problems quickly and troubleshoot them? Can it be used to provide APIs or with better observability(e.g. quantify that the system performance coming down is not sustainable)?
Another thing that is weird about this is the human aspect. Specifically role of the product manager. Ideally he should be someone who is experienced enough and acts as a moderator between what code/features go in and what don't. The client always wants the moon and a good PM's other real job is to manage this friction. Is he not able to push back? He is also the one that is looking at the project timeline and making such decisions. If that role is not being played by someone, then something is obviously wrong with the relationship.
I've had project that tried a similar thing, they tried to replace their team with cheaper coders. They went slower because of code slop and their product failed to launch on time and be successful.
The exchanges are bending head over heels to accommodate these IPOs[1] and make our retirement index funds the exit-liquidity strategy to the thievery of pump and dump actors that buy it low and then sell high? As i understand the way thievery works is:
1. List at many multiples of market valuation on an exchange. So if you company is just 10 billion$ nasdaq and theives collude and say "can make it 100 billion..".
2. Lots of institutional investors and rich billionaires get stock options.
3. All market weighted index funds — aka all *your* low expense ratio ETF money — have to re-balance and buy them, raising their value: the exit-liquidity event
4. Rich A**** get richer by making an profit by selling higher.
[1] https://www.economist.com/leaders/2026/03/31/index-providers...
I have kind of lost count of how many content creators have said personally to me traffic is meaningfully down because of all these chatbots. The latest example is this poor but standup guy: moneyfortherestofus.com.
This could be generally useful for a lot of algorithms like kmeans(ish) like Cut algorithm that projects points into eigen space — as opposed to the kernel space in Kmeans — need the affinity matrix(distance for K points).
I have my instagram, x on a locked down browser in a container with a fake profile that an LLM drives and finds the posts for specific users and compiles a gist of all the important things in my locality(or what u care about) every evening, without me ever going near that FOMO driven dumpster fire of tiktok/insta/x.
Best LLM RoI I made.
Your extension can and will likely read all your data and likely has broad permissions to just fax all your Gmail to remote cloud server.
Like this is a big deal to get a project to a state where volunteers are spun up and actively breaking tasks and getting work done, no? It's a python JIT something I know next to nothing about — as do most application developers — which tells one how difficult this must have been.
---
So, ultimately, to the question, what exactly is Sarvam AI? Is it a company that builds LLMs cheaply and open-sources them? Is it India’s Deepseek? Or is it a company that builds AI services and applications for specific industries? Like, say, Scale AI? Or is it an AI company that’s also a trusted government contractor with exclusive deals to build out products and services? Like India’s Palantir? Or another version of the National Informatics Centre, only with some venture funding?
---
They will have very narrow to zero understanding — don't need it to fix — of shear forces, navier stokes.
They will command high rates if labor is limited(a plumber in Indonesia will commande lower ppp adjusted hourly rates than America). CS education become a subset of applied math since graduate hiring of code-plumber will require a narrower certificate to fix an AI system — which works very much like how plumber working to fix a building leak is different from a person fixing a water pipe burst under a road.
A few AI systems will become dominant, That should be a mix of Anthropics and your Googles. They will hire code plumbers to plumb together all the things they provide.
You don't have to use much brain at all as a code-plumber. You become a remote journeyman logging in and plumbing with given tools, making sure there is low back pressure(a term where load on future plumbers interacting/fixing with ai decreases) and the like.
I've used CC with chrome to access social feeds autonomously and give me a notification at set times of the day, summarizing everything that I feel is worth knowing about— local events, local municipal announcements, some misc comics and some harmless fun with dogs/OnlyInNYC. It really takes out all the FOMO driven brain fuckery of feed and ads.
Or In other words you can get to decide two ways to use a lucrative property:
1. designate it private and draft usage of how you allow to use it, per your value system(as long as values don't violate any laws)
2. In face of competition, give up some values and agree to a legal definition of use that favors you.