For those not keeping up, I'm developing V.E.L.O.C.I.T.Y. IDE currently. Built from the ground up to be the definitive Agentic IDE. Last week, I had a MVP, this week I'm 50k LOC deep in rust, building it out, optimizing and added 2 massive features.
a Rust-based native browser built for agentic use. I had build MCP-Lite, which was incredible, along with it's full featured counterpart, MCP-Heavy, to anyone who wants to save a dime, you'll be happy to know, I've fully ported it in... Though I am sick and tired of Chrome eating memory, so I've been hard at work building a Rust-native browser from scratch. And by from scratch, I mean no CDP, REAL from scratch, not just a chromium port.
To the people who read shelved projects, Windows Automata is making a comeback! What started as stripping the sandbox (which worked exceptionally well btw), turned into a full integration now. At this point, the IDE is turning more into an all-rounder than anything else.
And then there's a few that you probably wouldnt know about...
ORC - Orchestrative Version Control, think Git, but built for multi-agent workflows, handling discourse and merge conflicts at write-time, instead of merge time. This will serve as both the internal state-controller, as well as an optional VC integration into Github, given I imagine none of you want to leave the comfort of your caves.
V.E.L.O.C.I.T.Y. Share, my file transfer protocol that hit within 90th percentile of wire-speed, will be the backbone for ORC and will be integral to the Rust-native browser integrated into the IDE to push even more performance out of it. It will also be opened up for any file transfers you need to do securely and for backing up wherever you'd like... Just alot faster and more securely.
V.E.L.O.C.I.T.Y. MCP - the Rust-based MCP that executes in nanosecond scale, instead of milliseconds, while completely stripping JSON and Node.js. This is the core-infrastructure of the entire thing. The models use it for whatever they want to do, be it browser, WA, or development, all tooling is natively faster than any MCP you've ever used.
V.E.L.O.C.I.T.Y. OCR - A rust-based Spline driven OCR engine, will be integrated to accommodate first-pass reading of images and docs. Given it does tasks in milliseconds that take other pipelines 100+ seconds, I think it's fair to say it'll be quite useful, especially for token reduction.
V.E.L.O.C.I.T.Y. NDA - The bread and butter of it all, whether it's for the persistent SiteMap merkle root/changelog, tool definitions, or wherever else JSON would have been used, it's faster, more secure and deterministic triples, so LLMs understand it inherently.
V.E.L.O.C.I.T.Y. Remote - Secured P2P remote desktop, both in and out. That means if you're doing any kind of remote work, the combination of Remote and WA, means you can automate it. Or you know, interface with the IDE remotely on the go. All connections are driven by a secure handshake server, then transition to pure P2P, SHA 256 encrypted comms, built on the NDA standard for auditability, defensibility and reliability.
This IDE is turning into alot more than just an IDE, it's the culmination of months worth of projects, all getting slapped together into a unified product that makes life easier and cheaper. Whether it's remote IT work, Software development, or anything to do with apps or internet, it's capable and it's efficient. Zero-allocation is the mandate and it's strictly adhered to, so expect a minute fixed ram allocation. No more bloated Electron based glorified browser apps, Rust and NDA, so it's reliable, secure and effectively as close to bare-metal as you can get (until V.E.L.O.C.I.T.Y. OS is done and it gets ported to it...)
Sorry for not keeping the game series alive, though you'll be happy to know I'm about 20% of the way done with Dwarven Stronghold, think DF, except built entirely in Rust, so it actually runs decently, with a few more expansions, like the ability to switch to first/third person mode on a dwarf and take control. Never thought 100k LOC would ever be this inept... But it's getting there.
I'll post updates along the way on both as they progress. Both have grown by 150k LOC over the past week, but the 50k LOC each the past 2 days has been night and day in terms of capability.
Disclaimer: My AI is busy, so I wrote this by hand. Except the cover-image, that's NanoBanana...
Top comments (7)
One thing I've noticed is that a 50K LOC diff is usually a symptom, not the root problem. It often means multiple concernsโfeatures, refactors, fixes, and cleanupโhave been bundled into a single review. That dramatically increases cognitive load and makes it harder to reason about regressions. Even if every individual change is correct, the overall context becomes difficult to verify. Breaking work into reviewable slices with clear intent tends to improve review quality far more than simply reducing the line count. Curious how others decide when a change is "too big" to review effectively.
The 50k LOC is perfectly fine. It's due to the internal VC handling everything during expansion, then I just push to git occasionally. Unlike git where you need to commit and merge and manage, all of it happens in real-time, including conflicted file-edit resolution, so multi-agent workflows dont break or step on toes. So while it looks terrifying for most, behind the scenes it's a well-oiled machine. I took most of the backbone for it from my foundry, which has been operating flawlessly for over 3 months now, so it's pretty solid. Part of reasoning for the 50k LOCs, is that it is a good way to test whether the new VC system actually handles it well. And sofar, it hasnt skipped a beat. It's syncing with the sitemap in realtime, resolving issues by the time I get back from making coffee and running entirely autonomously without need for intervention.
That's a whole lot of V.E.L.O.C.I.T.Y. projects โ hurry up and release V.E.L.O.C.I.T.Y. ๐ โ automated fried chicken and food delivery!
Haha, true, it's because they're all built on the framework. I mean if you have something that can run at wall street speeds, while doing more work than they do, on consumer hardware instead of custom ASICs, wouldnt you reuse it everywhere?
There's a real-world problem though โ how do you get people to believe it can actually do all that? You need something people can actually see working, otherwise it's like gold buried in a mountain โ nobody knows it's there until you dig it out. But once you do, you also need to figure out how to protect it so nobody takes it from you.
That was the initial thought with making Messenger, Share and Remote. Purpose being as a pure 'look what it can do' showcase. But I think the IDE is the safer bet as a compiled binary, cuz it's something people would actually use off the bat. For people to use a messenger app, it needs a community, file-share is just a 1 off thing usually and remote desktop people have their prefered platform. Vs agentic IDEs, people switch up and test new ones regularly, depending on whose got a promo going, or offers it cheaper. It'd also be a good way to build a community for the 3 apps, cuz you'd use messenger to get updates and send commands efficiently from your phone to your dev station. You'd use remote to see what's going on and you'd use share to quickly move files to and from the dev station. 4 birds, 1 stone.
Waiting for your release โ make sure it supports multiple languages and opens up beta applications, haha ๐