Logan Yang

Copilot for @obsdmd won the best LLM integration 2024!!! 🔥🔥🔥 This community is amazing, thank you guys sooooo much for your support! ❤️‍🔥 And this is only the beginning. Please stay tuned for more in 2025! 😉

@Internethunts @themetafranz @joethei @obsdmd So why did you?

@themetafranz @joethei @obsdmd How do you know? They never shared any data regarding revenue

The Obsidian team is growing from three engineers to four engineers. Competitive SF salary. Fully remote, live anywhere. Apply below.

Another one joins the "training on your data unless opt out" party. In the latest video I talked about the privacy levels of providers. Most people won't care but for anyone who still cares about privacy, consider running your own AI stack.

free growth strategy: 1. keep improving little by little 2. stay 100% user-supported 3. watch VC-backed companies gradually destroy their product and alienate their users

@dhuynh95 @steipete Nice stuff! I took a quick look and seems each call would be 5s slower? "Each call spawns a fresh TUI process (~5s startup overhead)"

@steipete github.com/dhuynh95/claud… I built a project to allow third party use inside Claude with just subscription 🫡 It's a CLI that behaves like claude -p but actually launches regular Claude Code, simulates key types, listens for changes in the JSONL conv and streams them in the CLI.

Anthropic now blocks first-party harness use too 👀 claude -p --append-system-prompt 'A personal assistant running inside OpenClaw.' 'is clawd here?' → 400 Third-party apps now draw from your extra usage, not your plan limits. So yeah: bring your own coin 🪙🦞

after turboquant and qwen3.5-35b-a3b, i got curious: how realistic is it to use kv cache as a document store today? to have vectorless, RAG-less search. so i prefilled 258K out of 262K context window on L4 (a budget GPU popular in prod). ~99% of the slot is pre-computed and stored, users load it on the fly in ~1s. system prompt + query append to the end, generation takes ~3K tokens, enough for search. at 99% fill rate, decoding runs ~20 tps on L4. i prepared some ego datasets (jina papers, which i know best), plus popular novels in chinese and english. the results are actually pretty good. some hallucination, but most answers are solid and well-grounded. what's more interesting is the cost: ~$0.26/h on L4 spot. single LLM. no vector database, no embedding model, no workflow/pipeline engineering. using kv cache as document store is nothing new, like the old CAG paper. but with quantized kv cache and modern attention (hybrid SSM-attention, GQA, MQA, MLA), the economics are changing fast. if we solve cold-prefill speed and decoding speed, and budget GPU costs keep dropping, the future of search could be vectorless. radical, but possible.

Software horror: litellm PyPI supply chain attack. Simple `pip install litellm` was enough to exfiltrate SSH keys, AWS/GCP/Azure creds, Kubernetes configs, git credentials, env vars (all your API keys), shell history, crypto wallets, SSL private keys, CI/CD secrets, database passwords. LiteLLM itself has 97 million downloads per month which is already terrible, but much worse, the contagion spreads to any project that depends on litellm. For example, if you did `pip install dspy` (which depended on litellm>=1.64.0), you'd also be pwnd. Same for any other large project that depended on litellm. Afaict the poisoned version was up for only less than ~1 hour. The attack had a bug which led to its discovery - Callum McMahon was using an MCP plugin inside Cursor that pulled in litellm as a transitive dependency. When litellm 1.82.8 installed, their machine ran out of RAM and crashed. So if the attacker didn't vibe code this attack it could have been undetected for many days or weeks. Supply chain attacks like this are basically the scariest thing imaginable in modern software. Every time you install any depedency you could be pulling in a poisoned package anywhere deep inside its entire depedency tree. This is especially risky with large projects that might have lots and lots of dependencies. The credentials that do get stolen in each attack can then be used to take over more accounts and compromise more packages. Classical software engineering would have you believe that dependencies are good (we're building pyramids from bricks), but imo this has to be re-evaluated, and it's why I've been so growingly averse to them, preferring to use LLMs to "yoink" functionality when it's simple enough and possible.

Dan says he's got Qwen 3.5 397B-A17B - a 209GB on disk MoE model - running on an M3 Mac at ~5.7 tokens per second using only 5.5 GB of active memory (!) by quantizing and then streaming weights from SSD (at ~17GB/s), since MoE models only use a small subset of their weights for each token

@kepano Great stuff! Seems it works off of a Cloudflare worker? Curious how you work around the YouTube rate limits/IP blocks?

Defuddle now returns Youtube transcripts! Paste a YouTube link into defuddle.md to get a markdown transcript with timestamps, chapters, and pretty good diarization! ...or if you just want to read it, try the new Reader mode in Obsidian Web Clipper powered by Defuddle.

The only to-dos method that has worked for me: Every week I make a weekly note, and write my to-dos for the week. I may add more items to it during the week. If any items didn't get done I roll them over to the next weekly note or drop them. That's it.

@elvissun @nz_mrl Didn't anthropic ban openclaw from using subscription tokens via oauth?

@nz_mrl opus for openclaw, codex for coding

we are so back

Given an embedding vector, you can tell which model produced it. I trained a 0.8M transformer that fingerprints embedding models by reading raw float digits (vocab size: 15). Full end-to-end, zero feature engineering.

Pure MLX of UMAP, t-SNE, PaCMAP, TriMap, DREAMS, and CNE for Apple Silicon. Beyond the algorithms, I also built a circle-splatting renderer in MLX - scatter-add alpha blending on Metal, piped to H264 hardware encoding. So 70K points, 784 dimensions, from raw data to rendered video < 5s on M3 Ultra. pip install mlx-vis Paper: arxiv.org/abs/2603.04035 Code: github.com/hanxiao/mlx-vis

RT @william_cobb: I see a lot of people talking about how awesome Claude is with Obsidian, but you know what works even better? Obsidian Co…

The paper says the best way to manage AI context is to treat everything like a file system. Today, a model's knowledge sits in separate prompts, databases, tools, and logs, so context engineering pulls this into a coherent system. The paper proposes an agentic file system where every memory, tool, external source, and human note appears as a file in a shared space. A persistent context repository separates raw history, long term memory, and short lived scratchpads, so the model's prompt holds only the slice needed right now. Every access and transformation is logged with timestamps and provenance, giving a trail for how information, tools, and human feedback shaped an answer. Because large language models see only limited context each call and forget past ones, the architecture adds a constructor to shrink context, an updater to swap pieces, and an evaluator to check answers and update memory. All of this is implemented in the AIGNE framework, where agents remember past conversations and call services like GitHub through the same file style interface, turning scattered prompts into a reusable context layer. ---- Paper Link – arxiv. org/abs/2512.05470 Paper Title: "Everything is Context: Agentic File System Abstraction for Context Engineering"

Someone points me PaCMAP. 3-phase optimization makes for a much more dramatic convergence than UMAP. Both now pure MLX on Apple Silicon. Fashion-MNIST 70K in ~2s. github.com/hanxiao/pacmap… github.com/hanxiao/umap-m…

Obsidian finally has a headless solution, great for running claws 🦞

This is the way

ACP is the future. I celebrate the ACP plugin and QMD, these are great work, sorry for not making it clear in the beginning. Internally our small team have been cooking on these for a long time since last year, so we are huge believers in these approaches but we have a much bigger vision for them, not putting it too modestly.

No intention to be hostile and apologies if that's how my reply was perceived. I'm curious about your plugin and will give it a try, but it was important to me to respect and celebrate Agent Client and QMD, and it seemed to me that you are trying to push your plugin at their expense. Maybe I misunderstood what you meant to say in your post. You too could have done better by explicitly recognising the work of others before telling us how you are addressing whatever issues you've identified. Sorry, let's be friends, I promise to check out your plugin when you have the ACP version ready. 🤝

A few notes on agentic @obsdmd (about a week in): - This is amazing. So fast, so flexible. - Bases for everything. All my notes have a `type` and matching template and structure. Bases organise them, the agent knows how to create, query, and update them. - All notes are top-level and linked by properties, tags, or links. Folders only for attachments, templates, etc... - Agent Client plugin (need to install from github using BRAT) is fantastic. I have a split window editor/agent at all times. I use it with @opencode ACP. Kudos to @rainbow_0219 🙏 - @tobi's QMD is the perfect companion. I have it re-indexed regularly, the skill installed, and instructions for the agent to use it to search notes (remember to use `query` for hybrid semantic/keyword search). - The CLI is promising, but the fact that you can't run it (or sync or any of ther other functions) without having the GUI available is my biggest problem with Obsidian currently. - Vault syncing through Obsidian Sync for mobile access, but also via the Git plugin to GitHub, and via a cron job to my OpenClaw and OpenCode websservers. That means that I can easily work with the notes agentically, also without the Obsidian client UI.

@intellectronica @obsdmd I've deleted the tweet. It's your account, if you don't like it here it's your call. Have a good day.

Good will is celebrating the success of two awesome plugins that were mentioned as being great solutions, and mentioning your plugin as an additional option, not addressing nonexistent "issues" with these products. It may well be that your solution is better, and I'm willing to give it a try, but that just wasn't good behaviour and citizenship. As a plugin developer you must know how much effort goes into these and how important it is when we can celebrate the success of someone's efforts. You can advocate for your own product without talking down on others.