File over app for AI https://xitter.cf/t.co/vVu3KrixaW https://xitter.cf/t.co/Ok61BXZIW1 https://xitter.cf/t.co/7vOO03PmQ8 youtube: loganhallucinates
PNW
Joined November 2009
- Tweets3.6K
- Following622
- Followers2.2K
- Likes17.8K
Pinned Tweet
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! 😉
This is what settles the Obsidian vs. Notion debate. Full Stop.
Notion vs Obsidian 論争に決着が付きました
Grok Bot × Notionが勝者です
ネット環境さえあれば、いつでもどこでも自分の知識のデータベースを呼び出せるし、そこからコンテンツを作ったり、知識をBotが活用するなどができる
応用も幅広くできて、たとえば、自分の特定の知能を与えた拡張知能Botが大量にできるし、似たような知識を紐づける作業をRoutine化もできるし、情報収集も知能を更新できるような情報を集めるとかできそう
結局、コードベースという開発環境の外の世界（Outer loop）はGrok Botが強いので、コードベース外のナレッジベースもGrok Botで外界のコネクターに接続して管理するのが、良いループ設計だ
ここでのループ設計はOuter LoopとInner Loopの境界と接続の話だ
Launching the Google Docs for the Agent Era but OPEN SOURCE!
Ask your agent to turn a markdown note, report, dashboard or deck into an interactive page. Share it with the world, or keep it private to the people you invite.
Your agent can do everything you can in OpenArtifacts: sign up, publish, comment, reply, and change who has access. You just chat and it does all the work.
Claude Artifacts and ChatGPT keep your pages inside their closed ecosystem, limited to their own agents.
OpenArtifacts works with ANY agent. It is MIT-licensed, self-host it or use our hosted service out-of-the-box.
Try it now!
Where should I go after this?
An urgent notice from the Railway team
blog.railway.com/p/security-…
You can use grok bot + Obsidian to work better than Notion 100% and you still benefit from data ownership and file-over-app that comes from Obsidian. Grok bot is great, I use it every day with Obsidian where it uses miyo MCP to operate on my vaults
Notion vs Obsidian 論争に決着が付きました
Grok Bot × Notionが勝者です
ネット環境さえあれば、いつでもどこでも自分の知識のデータベースを呼び出せるし、そこからコンテンツを作ったり、知識をBotが活用するなどができる
応用も幅広くできて、たとえば、自分の特定の知能を与えた拡張知能Botが大量にできるし、似たような知識を紐づける作業をRoutine化もできるし、情報収集も知能を更新できるような情報を集めるとかできそう
結局、コードベースという開発環境の外の世界（Outer loop）はGrok Botが強いので、コードベース外のナレッジベースもGrok Botで外界のコネクターに接続して管理するのが、良いループ設計だ
ここでのループ設計はOuter LoopとInner Loopの境界と接続の話だ
Miyo helps you own all your AI history across all apps, too. It’s the file-over-app philosophy extension for Obsidian for all AI apps.
It pulls all your chat history from ChatGPT and Claude down as md files on your computer so 1. you own them forever 2. new AI apps like grok bot instantly understand all your old chats if you want them to
Automating note categorization and duplicate checks against existing notes.
the new jev model is insane..
it can check your AI's work and make decisions inside software for $0.042 per million input tokens, with free output
here are the first things you should use jev for:
1. second brain
2. content workflow
3. post analysis
4. SEO article review
typesafe reports up to 193.6× faster results and 444.6× lower costs than the LLMs in its workflow tests
you give jev information and specific questions. it returns choices, scores or probabilities that your software uses to decide what happens next
1. second brain
when a document, message or meeting note enters your second brain, give jev the content and your categories
identify what it is, classify the topic and check for duplicates against existing notes
your software checks required fields, saves the content with its source and verifies that it was saved correctly
works well with the karpathy LLM wiki framework
2. content workflow
give jev your draft, and then reference your anti slop rules and voice DNA file (which defines how you write)
it will check for generic phrasing, repeated points and differences from your voice. use previous content and its results to estimate performance potential
then send results to the writing agent for revision, compare the performance estimates with results after publishing
3. post analysis
jev can compare a draft with previous posts and their results to estimate how well it could perform with your audience
evaluate the hook, topic and format against defined criterias. and compare those scores with the performance after it goes live
4. SEO article review
give jev your article / page draft, a target keyword and the articles currently ranking for that keyword (in the top 10 SERP)
compare how well they answer the query, cover the topic and provide useful information. use that comparison to estimate your article's ranking potential
then your LLMs can revise weak sections before publishing, and compare the estimate with actual rankings
so many more usecases, we will see many new upgrades to previous concepts and workflows we´ve read about now
Reranking improved BM25's relevant-first results from 11/24 to 18/24, though failures on a separate conclusion-checking test kept it out of the pipeline.
I tested Jev on real outputs from our pipeline and their source documents (3,548 pages) to see if the hype translates into something useful for us. For now I'm not adding it.
The most promising part was retrieval. On 24 searches, BM25 alone put a relevant page first 11 times. With Jev reranking the same results it went up to 18, which is a real improvement.
Using it to check conclusions was less convincing. I gave it 32 hand picked examples where I already knew the answer. It raised 10 warnings and 3 of them were on answers that were correct. It also let through a "greater than or equal to" where the source says "greater than". And 3 of those 4 mistakes came with confidence above 0.8, so I can't just filter on confidence.
The problem with a warning is that something has to check it before acting on it. If I trust it blindly I end up changing things that were already right.
The whole test cost about two cents and answers come back in one second, so I understand why people are excited. But a cheap API doesn't help me much if I have to double check what it tells me.
It's a small test, but I think the hype is ahead of the evidence. For our use case it hasn't earned its place yet.
10–170× faster and 3–20× cheaper classification than Flash/DeepSeek baselines in these tests, with ties on relevance and duplicates but a large loss on crawl-root selection.
We tested Jev against the models we use inside Context7's parsing pipeline (Gemini Flash, DeepSeek).
5 classification tasks. Results:
- 3 ties: query relevance, duplicate detection, website suitability
- 1 win: page classification — 85% vs 56%
- 1 loss: crawl-root selection — 27% vs 93%
- 10-170x faster, 3-20x cheaper
Tagging a single page: Jev wins.
Reasoning about a whole site's structure: it doesn't.
Crawl-root selection: given a URL + site nav, pick which section to crawl.
Query relevance: is the user's question about this library or something else.
Duplicate detection: are two snippets the same example.
Website suitability: is this site technical docs worth indexing.
Page classification: does this doc page have code, API ref, or info.
One big use case for Jev is search and relevance.
I've curated some of the most popular ones here (a growing list). 🧵
Labeling 1,018 paper summaries for $0.08 at 256 ms median per paper.
I used Jev to classify 1,018 AI research papers.
The result: $0.08 total cost and 256ms median end-to-end latency per paper.
The pipeline was:
1. Summarize each paper with DeepSeek V4 Flash
2. Send the title + summary + 24 possible topics to Jev
3. Use Jev to classify each paper
4. Visualize everything on 1kpapers.com
The summaries cost $3.99 on @togethercompute. The classifications cost $0.08 on @typesafeai.
So for just over $4 of inference, I ended up with a pretty useful way to explore the top AI research papers from the past year.
I think this is where things are heading: different models for different parts of the workflow, instead of using one model for everything.
I’m running evals on the Jev classifications before replacing the current ones, but the site is already live: 1kpapers.com
Filtering Zillow listings by criteria the site's standard filters don't cover, such as architecture and renovation status
Jev can serve as better natural language search on websites.
It can scan thousands of Zillow listings and classify properties by things you can't normally filter for - e.g. architecture, renovation status, proximity to freeways.
This was done in <20 sec and costs $0.18 👇
A launcher that interprets “the pdf I just downloaded” in ~100 ms per keystroke, going beyond the aliases, fuzzy matching and usage habits that conventional launchers rank by.
Also have been playing with @typesafeai Jev, insane!
So many immediate use cases and new apps are possible. What a time to be a builder!
Sharing some experiments here starting with:
Keystroke oracle / predictive launcher:
Your launcher ranks by aliases, fuzzy match, and habit.
Jev reads intent: type "the pdf I just downloaded" and the newest PDF is already the top hit with a full confidence on every keystroke, in ~100 ms
OpenArtifacts is an open source alternative to Claude Artifacts that works with any agent.
It's MIT licensed and you can self-host it.
Use Codex, Claude Code, Hermes, Pi, OpenCode or whatever agent you prefer to publish beautiful pages anyone can read from anywhere.
Run `npx openartifacts@latest install` and ask your agent to “publish this to openartifacts”.
Your page is up with a link to share.
The CLI also lets your agent manage your account and published pages, so you can ask it to unshare a page when you want it taken down.
Private sharing, inline comments and a versioned diff view are coming next.
Check out our github and give us a star if you like it! openartifacts.ai
My Obsidian vault is my portable agent memory.
I've been trying different harnesses and agent apps across devices, and the recurring pain is bringing my context along.
I want the next agent to know what I'm working on and what the previous agents have already done.
So I always ask agents to save their memory to my vault, including what they did and where they left off.
When I switch apps, I tell the new agent to read the relevant notes before getting to work.
That way, trying a new harness doesn't have to mean explaining everything from scratch.