Rachit (@rachitcodes) header
Rachit (@rachitcodes) profile picture

engineer. @atlanhq building @crosmoslabs, prev @localhosthq absorb all you can

bengaluru
Joined December 2021
Photo by Rachit (@rachitcodes)Photo by Rachit (@rachitcodes)Photo by Rachit (@rachitcodes)Photo by Rachit (@rachitcodes)Photo by Rachit (@rachitcodes)Photo by Rachit (@rachitcodes)Photo by Rachit (@rachitcodes)Photo by Rachit (@rachitcodes)Photo by Rachit (@rachitcodes)
Rachit retweeted
i never cuda expected this
60
553
32
9,195
170,445
Rachit retweeted
We’re excited to launch Atlan Trident 🔱 The shared context workspace for humans and agents. One place for every skill, agent, metric and ontology. Shared across your team, in the tools they already use. Better with every run.
10
19
2
41
2,053
Rachit retweeted
hey guys ran a small grocery platform for a bit, started from helping a local vendor and pitching them a solution for their problems. 45 customers. ₹10,981 revenue. Six weeks. I built it. Ran it. Supported it. Here's what actually mattered.
5
5
5
316
Planning events gets exhausting. They're making human connection a little easier. Plan, invite, and host events all in one place made effortless by Invytt.
12
10
5
69
7,040
Rachit retweeted
I was digging around on X and found Crosmos. It's a conceptual neighbour to what I'm researching, not as a vector backend but as a persistent context substrate for agents. Its core claim is that agents fail because context is scattered, stateless, and temporally flat. Their answer is a monotonic temporal knowledge graph, hybrid retrieval, contextual memories, soft forgetting, clusters, MCP tools, and editor skills. Their relevance is the architectural pattern, since context should be structured, persistent, queryable, provenance-aware, and exposed to agents (through MCP/skills). This supports my theory, we need a layer between agents and repositories, rather than another standalone agent framework. The most useful design lesson is their rejection of over-splitting. The “atomic” units should not become tiny disconnected fragments. They need to remain self-contained enough for agents to act on, while still being precise enough for retrieval and coverage analysis. Crosmos is broader organizational memory. I intend to stay narrower and sharper with repository structure, changed paths, context packs, branch drift, stale docs, trial artifacts, execution handoff, and evidence-aware coding context. Crosmos validates the market direction for me; my research's differentiation is codebase-native structural context with verifiable and measurable agent-execution outcomes.
People working around context infra and memory should realize, it's not about recall. Anyone can make their own custom RAG pipeline / filesystem setup that's good enough and works for them in a day. You're just adding 3 more steps to it. The main objective of a context infra should be how well are you maintaining the context over long running agents / noisy data / how well you handle contradictions / how well you forget. Yes "forget". Context is not just about storing a record. It's about giving the agent the intelligence it needs. How well you represent the context is how good your infra is. We're building crosmos.dev on the same primitives. Check it out and give your agents persistent context. Building at @localhostIND
1
3
7
407
momentum is built when you stop restarting and start improving the same thing
1
1
8
419
self launch video built in one day everything is ai dreamlike-creative-direction… for mobile: view horizontally double tap/click and enjoy the music @dreamlikedawar @swapnilsdream
12
9
40
3,262
the biggest moat i have is that i actually love what i do
7
3
27
1,369
the input interface has been the same for decades. with ai, software can now reason and act on your behalf but the interface is the bottleneck. why do i have to check sushi on 10 restaurants across 3 apps? why can't i just do it with a flick of a finger?! the world's about to get a new interface @agi_interfaces
94
76
27
305
89,350
kicked off the founders program today. 30 founders in the room, chosen from 1k who applied.
introducing localhost founders program: - 20 spots - irl in bengaluru - $300,000+ community fund - $100,000+ compute credits - workspace at our hq - access to hardware lab, software credits - incorporation, legal and media support - free flight from anywhere in india apply now. link in comments. applications close 30th march 2026.
25
23
4
380
41,092
Rachit retweeted
most people operate on a model of gain, it's almost universal. their usual thought patterns revolve around questions like what do i get out of this? what do i win? what's in it for me, to make this move, start this thing, etc? i think the inversion is more interesting & way more honest. my operational model is closer to nothing to lose. especially when you're building a company from zero, you're operating in open territory, or even interacting with anyone new. the downside is almost always capped. the upside is infinite. i guess some ppl might see this as optimism framed another way but i think of it as pure math. & once you internalize that asymmetry, it becomes a filter for everything. it tells you what actually matters & what's just fear. it will free you to take action.
39
47
3
540
33,969
The ability to ask the right question is more important than ever now. With AI generating slop and enabling you to build almost anything, having clarity on what to build, why to build it, who it is for, and what problem it actually solves is more important than ever. We are going to see a flurry of things people build that no one asked for. Critical thinking is the only real edge left - the real moat.
25
22
3
446
15,815
Judging by my tl there is a growing gap in understanding of AI capability. The first issue I think is around recency and tier of use. I think a lot of people tried the free tier of ChatGPT somewhere last year and allowed it to inform their views on AI a little too much. This is a group of reactions laughing at various quirks of the models, hallucinations, etc. Yes I also saw the viral videos of OpenAI's Advanced Voice mode fumbling simple queries like "should I drive or walk to the carwash". The thing is that these free and old/deprecated models don't reflect the capability in the latest round of state of the art agentic models of this year, especially OpenAI Codex and Claude Code. But that brings me to the second issue. Even if people paid $200/month to use the state of the art models, a lot of the capabilities are relatively "peaky" in highly technical areas. Typical queries around search, writing, advice, etc. are *not* the domain that has made the most noticeable and dramatic strides in capability. Partly, this is due to the technical details of reinforcement learning and its use of verifiable rewards. But partly, it's also because these use cases are not sufficiently prioritized by the companies in their hillclimbing because they don't lead to as much $$$ value. The goldmines are elsewhere, and the focus comes along. So that brings me to the second group of people, who *both* 1) pay for and use the state of the art frontier agentic models (OpenAI Codex / Claude Code) and 2) do so professionally in technical domains like programming, math and research. This group of people is subject to the highest amount of "AI Psychosis" because the recent improvements in these domains as of this year have been nothing short of staggering. When you hand a computer terminal to one of these models, you can now watch them melt programming problems that you'd normally expect to take days/weeks of work. It's this second group of people that assigns a much greater gravity to the capabilities, their slope, and various cyber-related repercussions. TLDR the people in these two groups are speaking past each other. It really is simultaneously the case that OpenAI's free and I think slightly orphaned (?) "Advanced Voice Mode" will fumble the dumbest questions in your Instagram's reels and *at the same time*, OpenAI's highest-tier and paid Codex model will go off for 1 hour to coherently restructure an entire code base, or find and exploit vulnerabilities in computer systems. This part really works and has made dramatic strides because 2 properties: 1) these domains offer explicit reward functions that are verifiable meaning they are easily amenable to reinforcement learning training (e.g. unit tests passed yes or no, in contrast to writing, which is much harder to explicitly judge), but also 2) they are a lot more valuable in b2b settings, meaning that the biggest fraction of the team is focused on improving them. So here we are.
The degree to which you are awed by AI is perfectly correlated with how much you use AI to code.
1,262
2,703
717
22,758
7,649,593
Rachit retweeted
reinforcement learning algorithms are so incredibly immature, it is a miracle that it is working at all. there is *so* much room for improvement on algorithms, parallelization and optimization for speed of learning. so, so, so much
22
15
1
548
24,676
Rachit retweeted
Computer use? More like “computer, use up all my rate limits in 2 minutes despite me paying $200/month”
Computer use is now in Claude Code. Claude can open your apps, click through your UI, and test what it built, right from the CLI. Now in research preview on Pro and Max plans.
135
63
12
2,671
117,287
Harsh truth: Your inability to focus on one thing for an extended amount of time will hurt you far greater than your lack of intellect.
56
81
5
717
13,276
OpenCode is about to get more powerful with remote sandboxes I showed a brief demo before, but here's a much more in-depth demo. it's not hard to add basic support for a remote env, but handling all the edge cases like when a remote env gets deleted is difficult. especially if care about good UX You never want to lose session data. so the choices are: run the session in your env, but run all tool calls remotely. that's too complex and painful. The other way is to just let the full session run remotely, but sync back all the session data in your env. We chose this path: we built a syncing system which logs all events in a way that we can always recreate your entire session. That means the remote env could get destroyed, but we can easily restore it. it also opens up other interesting ideas which we'll be exploring
78
89
25
1,406
305,660