Do we really need another vibe-coded LLM context/memory startup?
Do the authors have any benchmarks or test to show that this genuinely improved outputs?
I have tried probably 10-20 other open source projects and closed source projects purporting to improve Claude Code with memory/context, and still to this date, nothing works better than simply keeping my own library of markdown files for each project specification, markdown files for decisions made etc, and then explicitly telling Claude Code to review x,y,z markdown files.
I would also suggest to the founders, don't found a startup based on improving context for Claude Code, why? Because this is the number 1 thing the Claude Code developers are working on too, and it's clearly getting better and better with every release.
So not only are you competing with like 20+ other startups and 20+ other open-source projects, you are competing with Anthropic too.
This. Exactly this. Even relatively well working tools (from my experience and for my project types) like Agent OS are no guarantee, that Claude will not go on a tangent, use the "memory files" the framework tells it to use.
And I agree with your sentiment, that this is a "business field" that will get eaten by the next generations of base models getting better.
I mostly agree with this, if the goal were “better persistent memory inside Claude Code,” that wouldn’t be very interesting.
For a single agent and a single tool, keeping project specs and decisions in markdown and explicitly pointing the model at them works well. We do that too.
What we’re focused on is a different boundary: memory that isn’t owned by a specific agent or tool.
Once you start switching between tools (Claude, Codex, Cursor, etc.), or running multiple agents in parallel, markdown stops being “the memory” and becomes a coordination mechanism you have to keep in sync manually. Context created in one place doesn’t naturally flow to another, and you end up re-establishing state rather than accumulating it.
That’s why we're not thinking about this as "improving Claude Code”. We’re interested in the layer above that: a shared, external memory that can be plugged into any other model and tools, that any agent can read from or write to, and that can be selectively shared with collaborators. Context created in Claude can be reused in Codex, Manus, Cursor, or other agents from collaborators - and vice versa.
If one already built and is using one agent in one tool and is happy with markdown, they probably don’t need this. The value shows up once agents are treated as interchangeable workers and context needs to move across tools and people without being re-explained each time.
If markdown in a git repository isn’t good enough for collaboration, then why would any plugged in abstraction be better?
You imply you have a solution for current wholistic state. For this you would need a solution for context decay and relevant curation — with benchmarks that prove it is also more valuable than constant rediscovery (for quality and cost).
That narrative becomes harsher once you pivot to “general purpose agents” because you’re then competing with every existing knowledge work platform. So you’ll shift into “unified context for all your KW platforms” - where presumably the agents already have access (Claude today can basically go scrape all knowledge from anywhere).
So then it becomes an offering of “current state” in complex human processes and this is a concept I’m not sure any technology can capture; whether it’s across codebases (which for humans we settled on git) and especially not general working scenarios.
And I guess this is where it becomes a unified multi-agent wholistic state capture. Ambitious and fun problem.
right. I stopped reading at "ENSUE_API_KEY | Required. Get one at [dashboard](link to startup showing this is an ad)"
First thought: why do I need an API key for what can be local markdown files. Make contents of CLAUDE.md be "Refer to ROBOTS.md" and you've got yourself a multi-model solution.
Main objection to corporate AI uptake is what are you gonna do with our data. The value prop over local markdown files here is not at all clear to even begin asking that question.
It blows my mind that anyone can consider Stranger Things to be great anything. It's utter dross. It's like our standards have dropped massively over the last 50 years in almost every way, in literature, music, journalism, politics, movies, and TV.
Yeah it feels slow and laggy to me too and I'm not on an old laptop. Running on a M3 Macbook Pro here. I definitely notice the difference between using something like Ghostty (Rust based - super fast) and Toad (Python).
It's obviously way slower though. Also the point stands, it's written in a low-level, performance-oriented language. The author of Toad could have written it in Rust, Zig, C++, etc, but chose Python instead. He valued ease of development versus performance and the result is we get a laggy terminal.
I know for a fact that Textual can generate an entire frame in less than a 60th of a second. Any lag you see has nothing to do with the choice of language. A TUI just doesn’t require that much number crunching to use a low level language.
I’d be interesting in knowing what platform and terminal you observed the lag, when testing Toad.
Vim has a terrible user experience though. There's a reason everyone stopped using it as soon as they possibly could and moved to other text editors. Now the only vim users are the 60 year old+ greybeards who try to convince everyone they're such morons for not using it.
Stop trying to convince people to use vim, it sucks, it's got a terrible ux, it's not intuitive, it's overly complicated, hard to learn, arcane, and looks like ass.
I disagree, but I'm a 60 year old+ greybeard who has managed to get a bunch of other devs addicted to vim. My real goal is to keep the key bindings popular enough that I won't have to reprogram my muscle memory before I shuffle off.
Government IS inefficient though, and it's inefficient because there is zero competition and also complete job security. It's also inefficient because the employees are generally bottom of the barrel folks due to the incredibly poor wages.
So you can get people working in the government who couldn't get a job in the private sector if they tried, working with total job security (they can't get fired) for an entity with zero competition so there is no drive or motivation to get better or otherwise improve.
Whereas with private companies you can get hired quickly and fired quickly, meaning you have to perform well (motivation), you are paid better so you attract higher quality candidates, and also if the company does badly you go bankrupt, which means the whole company performs better or dies. The companies which remain win the market and are more efficient (as they are the companies which survived).
This is the fairy tale as it’s often told. Doesn’t match up with my experience. The incompetence and waste in private organizations is staggering. The free market efficiency and competency of the free market is greatly overstated.
That's a very US-centric viewpoint though, it doesn't apply to every government or society.
If the US government is more inefficient than others then there's something to be said about how it works, how it could be improved, instead there's only this rhetoric that doesn't invite at all the discussion about what are its failures and paths to improve, just recycled catchphrases supported by a cliché.
Private companies are also inefficient in many ways even with competition, why is that if competition is supposed to make inefficient companies uncompetitive? Maybe there's something else to discuss rather than these thought-terminating clichés...
Blow made his own language because he's so eye-wateringly arrogant and thinks every language (that he didn't make) sucks, and only he is smart enough to design a better language for programming games.
Seriously, this is why he did it. His ego and arrogance is off the charts and if it wasn't made by him, he thinks it sucks (e.g. he doesn't like Linux, probably because he realizes Torvalds is actually smarter than him). He also doesn't like C++ or Rust, again, it's probably a good indicator he has a deep inferiority complex and so he has to prove he's the smartest person in the world by writing his own, "better" language.
I.e. I don't think he's making a programming language for some "love of creating", I think he's doing it because he has a deep psychological issue/insecurity, which drives his need to always be the "smartest person in the room", his arrogance, the way he dismisses others who don't agree with his viewpoints etc.
Even if you don't like Jon, calling Jai an exercise in arrogance is simply untrue. When he started making Jai in ~2014, there were very few viable alternatives to C/C++ in the systems programming space that offered the kind of expressive power becoming of a langauge built this century. Rust is great, but it prioritising correctness is not always the right choice, especially not for games. Jai introduced many ideas that languages like Zig and Odin ended up adopting.
It may not have a public* release but, over the last decade (starting pre-Zig/Odin), Blow has discussed it extensively in his videos[0], enough that even ~10y was possible for someone to make a toy independent implementation[1].
Still then, it's a stretch to say that Jai influenced other languages. How could it when only a handful of game-centered applications have been built by a handfull of devs?
Rust and Zig developed features by cutting their teeth on large amounts of real software, not by following one guy's personal project that has no source, no library, no spec available.
Jai, odin and zig's creators are all part of the handmade network, a community of programmers. You are vastly underestimating blow's reach/influence.
Odin's creator has credited Jai as an influence. You can see him in the comments of old jai youtube videos (videos that go into a lot of depth about the language design). Odin's syntax and features are very similar to Jai, the influence is pretty clear. Odin has other influences of course but you could say it's "jai but open source".
Lastly, jai is not open source but it doesn't mean it's not available. You can message blow to get access to it. Many programmers have used it. There are third party jai libraries on github.
I've never heard of Odin or seen any projects written in it, seen a company hire for it, or seen it discussed at a PL conference. There's no stable compiler for it, and no spec. Yeah, I'm just one person, so maybe I'm just in my own bubble, but these are hobby projects with a very small communities.
> Still then, it's a stretch to say that Jai influenced other languages. How could it when only a handful of game-centered applications have been built by a handfull of devs?
Lots of people have seen his talks about the language, so why do you think its impossible it influenced other languages?
It's unlikely that the Rust and Zig devs are looking at one guy's gamedev focused vlog compared to feedback from tens of thousands of engineers writing tens of thousands of public projects in Rust and Zig.
Have they heard of Jai? Yeah probably. But it's barely a drop in the bucket as far as the PL design community goes.
Oh, yes, the Rust team does "market research" and interviews people to see how they use the language, where the pain points are, etc. They have talks at Rustconf about how they gather information on how the language is used. Never seen them mention Jai.
> How has Jai introduce ideas if it’s not even released?
These are orthogonal concepts. Jai can or cannot introduce ideas, and Jai can or cannot be released. As of now, it is in fact so that Jai has introduced ideas, and has been released to a closed group of beta testers.
> How can we claim to know what it did “right” when only a few projects have been built in it?
To judge whether Jai did something right, in my opinion, it suffices to read the documentation and experience someone else programming second-hand and take advantage of its offerings, namely making programming less tedious, more enjoyable, more safe. It appears to me that you set the bar of usefulness or success too high for no good reason.
I've watched enough hours of his streams to know that this is NOT a reductive take. Blow is one of the most arrogant developers and game designers, and believes that nearly everyone else is an idiot.
He's somewhat Musk adjacent in his need to be viewed as smart (but I guess he does so least have way more programming chops than Musk, so I'll give him that).
C++ & Linux are world-changing tools, but C++ & Linux really do suck in ways that become more offensive with taste. Rust makes very different tradeoffs than ones gamedevs want.
Regardless, if arrogance drives people to make new tools then we should be grateful for that arrogance.
Think it's more along the lines of Jon having the ability to create a language, and upon being dissatisfied with what he was using, decided to make his own.
GitHub is littered with pet languages that people have made, and doubt their reasons are simply about being "eye-wateringly arrogant".
Moving past that, people paying attention or wanting to use the language, usually means it appeals to them. Jai has fans and supporters, because they are able to look past or are not concerned about his personality quirks, but are focused on the quality and usefulness of the software produced.
I think you're projecting a lot of your own complexes and insecurities.
He built a language for a very specific task: building games. There were quite a few requirements for such a language. Opinionated? Yes. But that's how you get new languages: by having opinions. Along the way he changed the design and the assumptions several times (e.g. built-in SOA structures are gone) while keeping the original goal in mind and using it to build a custom engine and a game while building the language (thus validating the choices made).
If/when Jai is released hopefully sometime next year, I do hope the documentation includes the rationale because he talked a lot about why other languages don't cut it in his opinion in the early days of development.
Eh, I can write that comment because it's fairly easy to see this side of JBlow if you've been following his work for a while. He is so naturally abrasive about other people's work, loves shitting on things he didn't make himself, loves being the smartest guy in the room, and also is a covid is a hoax, anti-vaxxer, Trump supporter etc.
I don't think I'm the smartest guy in the room, and that's OK. I realised a long time ago that ego/arrogance isn't a great quality and it's far better to have a strong network of friends and supporters, and that doesn't happen if you're an arrogant prick.
And yes, he built the language (which is totally un-needed) because the "idiots" who made all the existing languages, didn't make one as good as in JBlows brain. Despite the fact that there are 1000s of games which are far better than anything JB has made written in C#, C++, Java, Rust, etc. Did Larian need to write a new programming language to make Baldurs Gate 3?
Only JB is arrogant to think that only a new language is good enough for him to make a game with. A game that is just a modern spin on Sokoban and where he paid a bunch of other game devs to use their puzzles! You could write this shit in three.js and it wouldnt look or feel any differently.
+1 to all of this. I can no longer deal seriously with Blow's ideas, programming language, or games because he can't present any idea without being highly condescending and critical of just about everyone else. I'm glad I've never had to work for or with him, because he's the type of coworker or boss that constantly makes everyone's lives miserable.
Yes, you can do good things with shitty tools. And you could stop and say: this is enough. But then we would probably never have any programming languages at all.
Haskell exists because idiots that made existing languages didn't make one as good as in Philip Wadler's brain.
Go literally exists because idiots cannot use programming languages created by geniuses.
Rust exists because idiots who made all the existing languages didn't make one as good as in Graydon Hoare's brain. After all, all browsers on the market were built in C/C++, who is he to think that he could create a better/different language? Shut up and get on with the program.
C# exists because idiots who created other languages didn't create a language Microsoft wanted to control, and also weren't as good as the one Anders Hejlsberg's brain. After all, Java was already there.
Except Java exists only because who created other languages didn't create a language as good as the one in James Gosling's (and Mike Sheridan's and Patrick Naughton's) brain. Again, C/C++ had already been there, they could've used that.
Is Blow abrasive and shits on a lot of things? Of course. If you can't see past that to what he's actually doing with the language he's implementing, it's your problem.
> Only JB is arrogant to think that only a new language is good enough for him to make a game with.
Lol. I think this is the textbook definition of projection. He literally never said nor implied this in any way, shape, or form.
If anything, creating a new language set him back several years.
I've never had SQLite corrupt a database file, and given how widely it's used literally everywhere without reports of corruption, and the incredibly extensive testing methodology they use to ensure that, your issues seem very unlikely to have been SQLite's fault.
I've tested almost every LLM which will work on a modern iPhone and Apples models are universally terrible in comparison to almost every open-weights model, they're so bad it's a joke amongst devs who work in this space.
The only thing it's useful is super basic tasks like sentiment classification, summarization (sort of), or stuff like, "Does this message contain toxic/bad language, answer yes or no only".
The OP clearly didn't mean "hallucination" as a bug or error in the AI, in the way you're suggesting. Words can have many different meanings!
You can easily say, Johnny had some wild hallucinations about a future where Elon Musk ruled the world. It just means it was some wild speculative thinking. I read this title in this sense of the world.
Not everything has to be nit-picked or overanalysed. This is an amusing article with an amusing title.
Do the authors have any benchmarks or test to show that this genuinely improved outputs?
I have tried probably 10-20 other open source projects and closed source projects purporting to improve Claude Code with memory/context, and still to this date, nothing works better than simply keeping my own library of markdown files for each project specification, markdown files for decisions made etc, and then explicitly telling Claude Code to review x,y,z markdown files.
I would also suggest to the founders, don't found a startup based on improving context for Claude Code, why? Because this is the number 1 thing the Claude Code developers are working on too, and it's clearly getting better and better with every release.
So not only are you competing with like 20+ other startups and 20+ other open-source projects, you are competing with Anthropic too.
reply