Mnexium

Curious to learn more. How well does file system work? At what scale are you operating? Do you see it scaling well into the future for your needs?

It seems local file system is the default for most folks starting out. I am curious to know when does this break down if ever.

Good write up - I think a lot of peoples journeys (at least my own) is completions api --> responses --> deeper into the tech.

How useful have you found the Agents SDK? Meaning the SDK taking care of sessions, chat history etc. is that very useful or just "its nice that it is there..." could have built it on your own if need be type of feature?

Much Luck on the EdTech tool - interesting on the PHP choice. Just personal preference?

Agreed, are you doing anything to take advantage of cacheing?

Anything else you can share about project?

I do see glimpses of lawsuits here and there on these guys. They all seem to just be flybys on the news to never appear again. Maybe something is happening and the resolution doesn't make the headline.

I think Facebook has been a shell of a useful company for a long time. Maybe some value in Instagram but I don't quite see it. It does seem like the AI companies are beginning to become unfavorable, with that being said, unsure how much of this is geopolitics where this is less of an issue with China and competing with China. I dont think I see the current administration/government doing anything.

Appreciate the message - I'd like to take the necessary steps to push my idea forward. In the ocean of everything AI, this idea is a single celled organism.

There for sure is something to be said about higher performers being high performers wherever/whatever role they're into. There is a large degree of luck but an even bigger degree of execution involved.

Part of the dilemma is would I be as successful financially if I went all in on W2 as well.

For me, I'd aim to be buddy#1 but I probably fall more inline with buddy#2.

I think the business partner advice makes sense and something I am considering. It is better to own 10% of 10M then 100% of 100K, but also just complimenting skill sets as well.

appreciate it

Very relevant, I think a lot of stuff in tech is romanticized, even things like doing a start up or getting into YC. We hear about the massive successes that the normal/median kind of disappears and is ignored.

I think from an odds perspective, one of the above is probably the highest likely hood if I leave w2.

Appreciate and grateful for your kind words :) Happy to hear of your success as an example of what makes this industry motivating.

You are correct my idea is infrastructure atop of current LLM technologies. Competitors exist and it may be that the LLM providers provide their own solution(s) as they have unlimited resources. I will be the first to admit that it is not "ground breaking" but rather boring but useful. I think personally if I have one success I can look at broader trends, but there is a real fear of the unknown.

I am curious on your opinion coming for government regulation just because it does not seem to be happening. I'd love to invest in the cloud companies (AWS, Azure, GCP etc) without being exposed to the broader business.

I have been conservative forever from a financial perspective. I feel comfortable enough to be out of W2 for a year or two.

but there is a real opportunity cost to leaving a w2 and jumping into an idea like this. Plus the fears that you stated as well are things I think about also.

Insanely difficult and borderline idiotic to ignore what you've stated. Difficult to know where we are on the curve and what will happen next.

I think there are a lot of similarities and parallels to those times happening now, maybe even more with more money and investment depending on it and expanding.

There are a lot of arguments to made on both sides. I guess the one i'd lean on is "Do we not build Amazon, Google and FB b/c dot com happened?" and the companies that I'm trying to build inspired by those guys.

I agree and appreciate it.

I am starting to figure out having a network and going into an idea would move things 10x or 100x faster, both from a success or failure route. You'd want to get to both as quickly as possible. I think what posting on HN gets me is some kind of signal toward something that I don't already have. A lot of generic but valuable advice here. Still validating in its own way.

I do agree with you - asking people for time is especially difficult "can you give me some advice" ... and then asking people for money is 100x more difficult. Even if they get something for said money.

Appreciate the post and thank you!

I think for me and the idea I saw a fundamental shift in the experience chatGPT provided once it was able to remember cross/post chat things. The APIs do not have this offering and are pretty bare, so being able to put this on top of any LLM API I think is a value add.

I am curious how you are using Claude in your case? Web-ui? Co-Work? Does not seem like API, but would help me.

> Honestly, getting VC cash does not look like the challenge it used to be. Having never jumped in this pool, don't know if I would sink or swim here - but it does seem to be the case that a good story will get funding. I do like to approach things from a practical angle, I don't necessarily need/want millions up front and a more conservative approach to find product-market-fit would be beneficial.

> 5 years - Cleaning pitch-deck vs slaving away? both sound pretty shitty to me :) I think there are pros/cons. Pitch deck route assume I can sustain myself for 5 years but it does come with an opportunity cost. It is ~1m earned w2 income. W2-slave route feels like a boring way to contribute to society/tech and maybe not reach full potential?

Appreciate the post - I do agree with your sentiment.

Not in big tech, but comfortable 250k in mid-teir expense remote suburb. The opportunity cost is real and something to truly be factored in. Something I can probably replace after this endeavor but it would require a lifestyle change (moving etc).

I think i'd be happy with the challenge of building a business if there was constant progress and the end of the tunnel was possible (not necessarily guaranteed, but not 0% either).

I think being acquihired by mid-level would be a success. If my path is posting clueless HN to being acquihired in 3-5 - i'd personally love that outcome and it would validate a lot of things (my ability to acheive this, idea validation etc).

Chops question is difficult to answer - with constant progress I have tons of internal motivation, but one "no" and I tend to feel like shit (a character flaw for this line of work imo).

I agree - not having a network to vet the idea is possibly the easiest part to have. It means I dont have the network for sales like you said, the network to raise funding, the network to build upon what I already have.

Not wrong there.

Appreciate the message. I think this is the most practical advice to follow until there is an event/time to fully disconnect from the W2.

I'd love to hear/learn more about your journey. From a 0-1 build mindset but also going from 1 or 0.75-to-first customer and beyond.

I think we see a lot Zucks and AirBnB guys, and with AI 100M ARR in 2 months type stories, but not so much tortoise approach stories.

I am naturally extremely conservative. Part of the dilemma is maybe seek promotions at work or stay at same level and build side project.

I think doing something like this is very much high risk/high reward. Part of the challenge (albeit a good one) is the salaries are fairly good - so this approaching that is not an overnight thing. From a personal circumstance side of things there is also that opportunity cost.

"Don't quit your W2 until you get some real traction on it." I think this is probably the best path.

Agree and fairly practical advice and one that comes up frequently when watching videos on yc founders etc.

I think another thing I'd like to point out is that it isn't necessarily 'quit your job' as it can also be 'do i take on more responsibility at work' or new role vs keep same role and split time between project and work.

Good advice - appreciate it

I agree and good advice. Being so new to this and not having a network to bounce ideas off of or even try the app makes that part difficult to go from 0 to 1 for me personally.

While I agree that competition exists and it is validating that the labs and others are working on solving this problem. I am trying to focus on a broader solution. Competitors will focus only on memory - but not so much on the other infrastructure requirements to create an AI app (chat history, etc).

I am also trying to create a moat around the difference in how we validate and create memories. Most will have a single pipeline, I want to have multiple and compare - this does increase my COGS however.

Appreciate the message -

I agree that users (hell even 1 real user) would be motivating and validating from an idea perspective.

May I ask how long it took you to find your first user? and the path you took to find said user(s)?

Makes sense.

For curiosity's sake, have you had it try to attempt captchas?

If so, what were the results?

I generally am curious too - Im running into a similar issue with my tool build.

Difficult just to get feedback. This seems like a cool project, I dont have a ticketing system otherwise i'd try it.

Curious - how does it perform with captchas and other "are you human" stuff on the web?

Would love to hear more.

New entrants being folks outside the model providers themselves? I do think the database layer can be decoupled from the memory/creation layer

Im probably less worried about "AI Powered autonomous weaponry" in the near term and think its probably easier to imagine more human-remote-controlled scenarios of robotics. Meaning todays drones in warfare are still flown by humans, can China or the US get to a state where humanoid robots are controlled by humans? I think thats believable and easier to achieve while being just as scary.

I’m working on Mnexium, a memory + context layer for AI apps.

It handles persistent chat history, long-term semantic memory, user profiles, agent state, and rolling summarization.

It sits between your app and the model and works across OpenAI, Claude, and Gemini — you can switch models without losing context and still maintain memories.

The goal is to stop every team from re-implementing slightly broken memory systems, and to make memory user-owned and portable instead of tied to a single model or vendor.

Currently live with JS/TS and Python SDKs. Looking for feedback if anyone has any - good idea, bad idea? etc.

https://mnexium.com