Kneeanderthul
@kneeanderthul
Followers
3K
Following
48K
Media
3K
Statuses
10K
Knuckle dragging around the world. Look nice, hit over head, take home.
Joined October 2021
The 🐐. Seeing this video is always sobering. Cheers to all 🥳.
4
2
20
This concept is astonishing to see in real time . Id conceptualized something similar so we could upcycle in the future. Help folks run local models using a mining rig style system. Ahmad making it happen, just goes to show the spirit of innovation. Truly inspiring!.
wanna see how i use local llms?. a thread of blogposts where i go over:. > my setup.> how local llms actually work.> why i use them & what i do with them
0
1
1
The most fun image & video creation tool in the world is here. Try it for free in the Grok App.
0
127
992
The Wild West to capture centralization for all things RAG is real . Right now I’d say Huggingface 🤗 has captured most items. But it’s always interesting to see how it it’ll all play out.
One of the issues for AI users is knowing which option is best for their specific problem/s. @Edenlayer is focused on enhancing discoverability of AI Agents and Apps, helping you find the right tool for the right job. Join our community for updates:.
0
0
1
The age of spam is among us . N8N and other AI tools will soon lead to more folks thinking they're "optimizing " their business. Once you're part of the spam, there ain't coming back . Don't join the spam, be more human.
I get ~10 spam calls per day (various automated voicemails, "loan pre-approval" etc) and ~5 spam messages per day (usually phishing). - I have AT&T Active Armor, all of the above still slips through. - All of the above is always from new, unique numbers so blocking doesn't work.
0
0
3
Once it breaks into mainstream culture. Will CULTURE as a whole care for the NFTs? Will they care about the physical items instead ? Or maybe something entirely unexpected. Its all very curious and I hope folk manage their expectations.
0
0
0
As the space matures, will this be the consensus or the echo chamber ? . Do I think Clayno, can become a successful IP brand. Absolutely . Would I consider owning an NFT the same as owning stock in a company ? Not at all.
It's almost comical how undervalued @Claynosaurz are at this point. A year ago you could argue they were still catching up to the other IP-focused brands. No longer the case. And a year from now, they'll be so far ahead. Being on solana isn't doing us any favors, but at some.
3
1
5
Outlines will supercharge custom semantic layers. These updates are exciting! More folks , in some distant future, will have access to meaningful local db with LLMs, in large part due to their efforts and others with similar goals. Congratulations 🎉.
Outlines is the universal interface to structured outputs, bridging the gap between the low-level interface of structured generation and the abstractions you are familiar with. We're very happy to announce that XGrammar is now available as a backend alongside outlines-core!
2
0
6
Ship ship ship ship ship . My goodness! How can I get a Pydantic hoodie or something 😅.
🚀 Durable Agents in Pydantic AI - Temporal support is now available. Run AI workflows that:.• Survive restarts, failures & deploys.• Pause for hours/days & resume exactly.• Handle async + human-in-the-loop reliably. Deterministic logic → workflows.I/O & model calls →.
1
0
5
Before @recallnet there was tableland. I distinctly remember raising tables, a precursor to structures saved data on polygon. Now Recall is launching their own chain, and truthfully, rigs still hold a special place for me. Just scooped another one up.
Minted my 1st NFT using @tableland__ . It's a great experience. We ran into technical difficulties but managed to work around them. It's pretty interesting stuff.
3
0
3
How ChatGPT “remembers” between chats 🧠💬. There’s a hidden file created as you work. As “important” info builds in any prompt, it’s saved to a Model Set Context (aka Memory Injection Block). Every new chat gets this file → giving the illusion of memory. Ask any chat: “What’s
1
0
1
Used Pydantic to store structured Json into db . Used Pydantic AI to avoid the pain brought by brittle promoting .=.Focus on what's important! Building with LLMs is getting easier 🥳.
Pydantic AI @pydantic has the best LLM-native design abstractions. It's a new Langchain/Langgraph 2.0!. We successfully migrated from our custom implementation and it significantly decreased all the boilerplate code (streaming, LLM APIs, data model, context management etc. ) we.
1
0
2
Mathesar built a pretty awesome all to mess feel like you can have a meaningful interaction with db at any level. I'm excited to learn from them and see what we share and how to improve users interactions . Amazing work from Mathesar Foundation.
🚀 Mathesar 0.3.0 is here, and we now support dark mode! Plus, we’ve refreshed the UI, added a way to install Mathesar using just one line, and made records easier to find and create. Learn more in our release notes: .
3
0
6
Got some pretty cool names!
Introducing Nounified ENS identities.↳ (yourname).⌐◨-◨.eth. Free to mint. Yours forever. The fastest way to make your wallet instantly recognizable in the ⌐◨-◨ universe. Mint now — link below & in bio ↴
1
0
3
🫡 a simple push to make it simple to run @pydantic AI locally. Maybe I was the only one in the struggle bus, but I hope this helps all who need it. That API-key and endpoint quirks were just getting to me. But we done did it. Thanks again to the team!. Semantic layers for
1
0
8
LLM readable documents is the wave. Having README with echo drops for additional problem scope is the next logical step and also shout out to @MozillaAI for their work. I'd attach Firefox to the name for so long I forgot Mozilla was the company 😅.
🔧 Callbacks module. ⚙️ Batch processing example. 🤖 LLM-readable docs. 📝 Full changelog. Tell us what you'd love to see next!.
2
0
4
LangExtract is nice!.Pair it with Audio/Video → Text models and you’ve got a powerful base. BUT THIS ALONE ≠ memory. Wrap their schema with metadata, connect it to retrieval, and watch it grow into an actual Mesh. RAG = RETRIEVING. p.s. run it locally, own your data.
✨Announcing LangExtract! ✨. Our new open-source Python library for information extraction, powered by #Gemini. ✅ Turn text into structured data .✅ Trace every insight to its source .✅ Visualize results instantly. Explore the blog by @AkshayGoelMD and Atilla:
1
0
4
[PDF Upload]. │. ▼.[Image-to-Text (OCR fallback) + Text Extraction]. │. ▼.[Chunking + Metadata Tagging]. │. ▼.[Per-Chunk Embedding/Analysis]. │. ▼.[Reasoning + Answer Assembly]. │. ▼.[User Response]. - All the best 🥳.
Anthropic dropped "native" PDF reading support in Claude Code 1.0.67. I used claude-trace to understand how that works. TL;DR: PDFs are uploaded to Anthropic's servers. We do not get to know what the converted plain text looks like, whether (and how) tables are extracted, etc.
0
0
0
This sounds oddly familiar . Releasing the robust version. Adding the semantic layer after .( @pydantic AI) . This will unlock EXACTLY what you’re imagining and more . 🥳
more than anything else, I want an LLM filesystem. I don’t want to *think* about files, how they are organized. give me a semantic system organized by concepts so I can ask things like:. "show me the pdf of that paper I used in my youtube video on PTX 6 months ago".
1
2
4
If @github and @obsdmd joined forces and ran locally on your device. (currently on @PostgreSQL). It's powered by an application-managed UUID at ingest, creating a universal key to track and manage across ANY system (dual db with @qdrant_engine🧠) . Accessible to all! . UI next
0
0
1