Sign in

qdrddr.bsky.social

@qdrddr.bsky.social
91 followers 121 following 580 posts

Claude Certified Architect targeterpro.com/linktr-ee-qdrddr

PostsRepliesMedia
qdrddr.bsky.social @qdrddr.bsky.social · 1h
Uzu fast #inference engine 4 #macOS & #iOS. Native #Rust. Bindings: #Python, #TypeScript, #Swift or #CLI. ~ x4 faster vs. llama.cpp & #MLX with #Qwen 3.5 9B Q4, on #M5 #MacBook Pro: - Uzu 92 tokens/s - llama.cpp 22 t/s - MLX 25 t/s Needs lalamo model format converter #AI #LLM #OpenSource MIT Lic
github.com
GitHub - trymirai/uzu: A high-performance inference engine for AI models
A high-performance inference engine for AI models. Contribute to trymirai/uzu development by creating an account on GitHub.
020
qdrddr.bsky.social @qdrddr.bsky.social · 2h
New #OCR models @LightOn: 0.8B, 1B, 4B. 1500 tokens per page, while generating slightly fewer tokens. #olmOCR #Bench with 0.8B achieved 85.5%; 4B - 86.3%. Competes with Mistral OCR 4.1, Chandra-OCR-2, Jina OCR-v1, dots.mocr, Infinity-Parser2-Pro. #AI #LLM #OpenSource Apache 2.0 lic
huggingface.co
LightOnOCR-3: High-Performance OCR and Layout Extraction in One Model
A Blog post by LightOn AI on Hugging Face
020
qdrddr.bsky.social @qdrddr.bsky.social · 07/10/2026
@anthropic.com released #Hiku 5.5. Competes with GPT Luna. After 100k tokens are priced 5x more. Input: $0.10 / $0.50 Output: $0.50 / $2.50 #LLM #AI www.anthropic.com/claude-haiku...
anthropic.com
Introducing Claude Haiku 5.5
Claude Haiku 5.5 is our fastest, most capable small model. Built for high-volume work like summarization, subagents, and browser use.
010
qdrddr.bsky.social @qdrddr.bsky.social · 07/10/2026
000
qdrddr.bsky.social @qdrddr.bsky.social · 07/10/2026
GitHub: github.com/qdrddr/clear-your-tools
github.com
GitHub - qdrddr/clear-your-tools: Cut input tokens by 30% while preserving LLM focus and pruning irrelevant MCP tools
Cut input tokens by 30% while preserving LLM focus and pruning irrelevant MCP tools - qdrddr/clear-your-tools
010
qdrddr.bsky.social @qdrddr.bsky.social · 07/10/2026
#Perplexity released 2 sizes of #Embedding #models: 0.6B & 9B. Text or Image. #OpenSource #MIT lic. Interestingly they operate in the same latent space: embed with 9B, retrieve with 0.6B. Late interaction #ColBERT, 128 dimensions. #AI #LLM #SLM
huggingface.co
pplx-embed-v2 - a perplexity-ai Collection
Collection of our second set of text embedding models
030
qdrddr.bsky.social @qdrddr.bsky.social · 07/10/2026
#Decisions #API is available on #OpenAI with gpt-6-luna model only. Text & Images. X10 faster than Responses API. Input costs $0.10 per 1M tokens. You pay only for input tokens: there are no cache-read, cache-write, or output-token charges. POST /v1/decisions #AI #LLM #Jev #SystemOne #GPT
developers.openai.com
Decisions | OpenAI API
Use the Decisions API to check conditions, select from fixed options, and score text and images against a rubric.
010
qdrddr.bsky.social @qdrddr.bsky.social · 06/10/2026
Hope to see its performance on MTEB leaderboard soon.
000
Reposted by @qdrddr.bsky.social
Unsloth AI @unsloth.ai · 06/10/2026
Google releases EmbeddingGemma 2, a new open model that runs locally on 0.5GB RAM. The 740M parameter embedding model combines a 270M text model with vision (170M) + audio (300M). Run & train the model via Unsloth. GGUF: huggingface.co/unsloth/embe... Guide: unsloth.ai/docs/models/...
3989
qdrddr.bsky.social @qdrddr.bsky.social · 06/10/2026
Open an issue or PR in the GitHub repo if you see a problem.
000
qdrddr.bsky.social @qdrddr.bsky.social · 06/10/2026
youtu.be/qe-lLylibfM?...
youtu.be
Clear Your Tools: Tier Manager, Saves 73% of input tokens
YouTube video by Di B
010
qdrddr.bsky.social @qdrddr.bsky.social · 06/10/2026
New ver. release! Manually toggling #MCP tools on/off in #Cursor? Typing “use XYZ tool”? Stop Drowning #Agent in MCP Noise. CYT is here to fix this - removes irrelevant tools, schema properties & enums, injects statisticaly grounded tools with tiering manager. 🔗👇🧵 #NoTelemetry #OpenSource Apache 2.0
medium.com
Clear Your Tools: Less Context, Lower Costs, and More Effective Coding Agents
The new version of Clear Your Tools (CYT) introduces an adaptive Tier Manager moving tools with agent signalling tool use. Makes your agent cheaper and smarter.
350
qdrddr.bsky.social @qdrddr.bsky.social · 03/10/2026
#PostgreSQL #in-process with #vectorsearch extensions directly #embedded in your #Python app. #OpenSource Apache 2.0 lic - Similarly search: #pgvector + #pgvectorscale, high-performance storage - #FullText search: #pg_textsearch, #BM25 and #ranking #AI #LLM #Embeddings #Agents #RAG
github.com
GitHub - Ladybug-Memory/pgembed: Embedded PostgreSQL for Agents
Embedded PostgreSQL for Agents. Contribute to Ladybug-Memory/pgembed development by creating an account on GitHub.
040
qdrddr.bsky.social @qdrddr.bsky.social · 26/09/2026
60 #opensource #Jev #alternatives #classifier #AI #Models #LLM huggingface.co/spaces/multi...
huggingface.co
Jev Decision Index - a Hugging Face Space by multimodalart
Benchmarks and news on various repros of TypeSafe's Jev
040
qdrddr.bsky.social @qdrddr.bsky.social · 24/09/2026
#Jev #opensource #alternative Apache 2.0 lic #AI #LLM #Model github.com/allebee/jevk5
github.com
GitHub - allebee/jevk5: JevK5: open-weight alternative to TypeSafe Jev. Typed decisions with probabilities in one forward pass; Apache-2.0 weights and code.
JevK5: open-weight alternative to TypeSafe Jev. Typed decisions with probabilities in one forward pass; Apache-2.0 weights and code. - allebee/jevk5
031
qdrddr.bsky.social @qdrddr.bsky.social · 23/09/2026
#LLM based #Jev alternative #AI #TypeSafe #Reranking #Model #OpenSource. Apache 2.0 lic
github.com
GitHub - nokia-applied-research/AnyJev: Turn any LLM into a Jev-style decision model: typed decisions, real probabilities, no training. (continue updating)
Turn any LLM into a Jev-style decision model: typed decisions, real probabilities, no training. (continue updating) - nokia-applied-research/AnyJev
031
qdrddr.bsky.social @qdrddr.bsky.social · 22/09/2026
An extremely fast self-hosted alternative to Jev built on the #ModernBERT #embedding #model. Laya is multilingual and #typesafe, #python. #OpenSource Apache 2.0 license. #AI #LLM #Reranking GitHub github.com/NandhaKishor...
150
qdrddr.bsky.social @qdrddr.bsky.social · 21/09/2026
#Claude #Code now reads AGENTS.md (when CLAUDE.md) is missing #AI #Agents #CodingAgent #LLM #ContextEngineering github.com/anthropics/c...
agents.md
AGENTS.md
AGENTS.md is a simple, open format for guiding coding agents. Think of it as a README for agents.
010
qdrddr.bsky.social @qdrddr.bsky.social · 15/09/2026
#Linter for #Tailwind design. Helps #agents to verify and build a better #UI. #OpenSource MIT lic #shadcn github.com/shadcn-ui/lint
github.com
GitHub - shadcn-ui/lint: An agent-first linter for Tailwind design systems. Write design system rules that agents can verify.
An agent-first linter for Tailwind design systems. Write design system rules that agents can verify. - shadcn-ui/lint
060
qdrddr.bsky.social @qdrddr.bsky.social · 13/09/2026
Pack #Codex, #Claude #Code, #Hermes, #PI #Agents into a container, consume through a single unified #harness protocol #API. #OpenSource Apache 2.0 lic #AI #LLM #CodeAgent github.com/HarnessRoute...
github.com
GitHub - HarnessRouter/harnessrouter: HarnessRouter Community Edition: the self-hosted, Apache-2.0 edition of the unified interface for agent harnesses. Run Codex, Claude Code, Hermes, PI, DSH, and mo...
HarnessRouter Community Edition: the self-hosted, Apache-2.0 edition of the unified interface for agent harnesses. Run Codex, Claude Code, Hermes, PI, DSH, and more through one API, with sessions, ...
040
qdrddr.bsky.social @qdrddr.bsky.social · 05/09/2026
How to game a popular opinion on #AI. #LLM #AGI #Agents #RAG medium.com/@qdrddr/how-...
010
qdrddr.bsky.social @qdrddr.bsky.social · 01/09/2026
Check this out, my first avatar video by @elevenlabs.io. Save tokens, prevent tool call hallucinations. Windows/macOS. #OpenSource Apache 2.0, free offline. What do you think? Thumbs 👍 up is highly appreciated :) #AI #LLM #Agents youtu.be/duj81kxfWCc
youtu.be
ClearYourTools: Cut MCP Tool Context by 10,000 Tokens & Reduce Hallucinations
YouTube video by Di B
040
qdrddr.bsky.social @qdrddr.bsky.social · 26/08/2026
#PDF to #Markdown convertor, accuracy and speed ahead of #MinerU and #Dockling. Hybrid mode occasionally uses #VLLM, but mostly #OCR only. #RAG #AI #OpenSource Apache 2.0 lic github.com/datalab-to/m...
github.com
GitHub - datalab-to/marker: Convert PDF to markdown + JSON quickly with high accuracy
Convert PDF to markdown + JSON quickly with high accuracy - datalab-to/marker
140
qdrddr.bsky.social @qdrddr.bsky.social · 25/08/2026
I believe it’s mostly more efficient caching.
010
qdrddr.bsky.social @qdrddr.bsky.social · 25/08/2026
Why do ppl omit #quantization level when presenting #LLM bench? Shy? 🙂 It degrades intelligence. Even state-of-the-art methods inevitably trade accuracy for efficiency. Compress a high-rez photo: much smaller, some detail is inevitably lost, and some become unrecognizable. There’s no free lunch #AI
010
qdrddr.bsky.social @qdrddr.bsky.social · 23/08/2026
#Inference engine to run #LLM locally. Claimed to achieve x2-x4 faster performance vs. #Ollama. #AI #OpenSource Apache 2.0 lic github.com/FlashML-org/...
github.com
GitHub - FlashML-org/FreeToken
Contribute to FlashML-org/FreeToken development by creating an account on GitHub.
150
qdrddr.bsky.social @qdrddr.bsky.social · 22/08/2026
Local #LLM proxy for #agents on #macOS. Switch between models and providers. Manage #API keys or #OAuth. #AI #Apple #OpenSource MIT lic github.com/nguyenphutro...
github.com
GitHub - nguyenphutrong/quotio: Stop juggling AI accounts. Quotio is a beautiful native macOS menu bar app that unifies your Claude, Gemini, OpenAI, Qwen, and Antigravity subscriptions – with real-tim...
Stop juggling AI accounts. Quotio is a beautiful native macOS menu bar app that unifies your Claude, Gemini, OpenAI, Qwen, and Antigravity subscriptions – with real-time quota tracking and smart au...
060
qdrddr.bsky.social @qdrddr.bsky.social · 14/08/2026
@openaibot.bsky.social GPT-5.6 Sol running on Cerebras at 750 tokens per seconds. #LLm #AI openai.com/index/previe...
openai.com
Previewing Ultrafast mode: GPT-5.6 Sol at up to 14X the speed
Preview Ultrafast, a new OpenAI API service tier that runs GPT-5.6 Sol up to 14× faster. Powered by Cerebras, it delivers up to 750 output tokens per second.
010
qdrddr.bsky.social @qdrddr.bsky.social · 14/08/2026
Happy to share that I’ve passed the Claude Certified Architect Foundations Exam. #ClaudeCertified #Claude #ClaudeCode #Anthropic #AI #LLM www.credly.com/users/damien...
020
qdrddr.bsky.social @qdrddr.bsky.social · 13/08/2026
Well, let’s call this agent a hard lock-in. And a soft lock-in would allow it to run elsewhere with some reasonable effort. For instance, you can run OpenCode or Pi everywhere, in every cloud, while you cannot run Cloudflare OS agent elsewhere. Does it make sense?
120
qdrddr.bsky.social @qdrddr.bsky.social · 11/08/2026
FP64 → FP32 → FP16/BF16 → FP8 → FP4. Why stop? Why not natively trained ternary neural nets with weights {-1, 0, +1}? Instead of multiplying → sign changes, zeroing, & addition, eliminating floating-point GPU HW from the core weight ops. Why we heading toward a post-FP era for Models so slowly?
020
qdrddr.bsky.social @qdrddr.bsky.social · 10/08/2026
My point was that it is Apache 2.0 lic technically, but you’ll not be able to run it elsewhere.
120
qdrddr.bsky.social @qdrddr.bsky.social · 10/08/2026
Well, if you run your own k8s, you are not locked in. And you can decide to run on Baremetal, or in any clod and then move across, no lock-in, but more ops.
120
qdrddr.bsky.social @qdrddr.bsky.social · 10/08/2026
#RDF #graph #visualization and #SPARQL query #OpenSource MIT lic #GraphDB
github.com
GitHub - kvistgaard/nodica: RDF graph visualisation with image-filled nodes
RDF graph visualisation with image-filled nodes. Contribute to kvistgaard/nodica development by creating an account on GitHub.
030
qdrddr.bsky.social @qdrddr.bsky.social · 10/08/2026
⭐️GitHub: github.com/marcelroed/g...
github.com
GitHub - marcelroed/gigatoken: Language model tokenization at GB/s
Language model tokenization at GB/s. Contribute to marcelroed/gigatoken development by creating an account on GitHub.
010
qdrddr.bsky.social @qdrddr.bsky.social · 10/08/2026
Extremely fast #tokenizer, x1000 faster than #Tiktoken. #Python & #Rust bindings. No WordPiece, not optimized SentencePiece. #Windows not tested. #OpenSource MIT lic. 🔗 in first 💬 👇
130
qdrddr.bsky.social @qdrddr.bsky.social · 10/08/2026
They have released a tool to classify whether a document can be extracted with or without a model. Separate tool bsky.app/profile/qdrd...
000
qdrddr.bsky.social @qdrddr.bsky.social · 10/08/2026
@cloudflare.social released a #harness #agent, confusingly called #Cloudflare OS, that includes an agent runtime, a workspace, and additional abstraction layers. Heavily dependent on Cloudflare’s own infrastructure. Beware of vendor lock-in in disguise. #AI #LLM #Agents #OpenSource Apache 2.0 lic
github.com
GitHub - cloudflare/cloudflare-os: Agent workspace built on Cloudflare Workers for creating documents, building apps, and running agents with your company’s context and systems.
Agent workspace built on Cloudflare Workers for creating documents, building apps, and running agents with your company’s context and systems. - cloudflare/cloudflare-os
160
qdrddr.bsky.social @qdrddr.bsky.social · 08/08/2026
pgBackRest - reliable @postgresql.org backup & restore #extension. #PostgreSQL #Backup #SQL #DB #OpenSource MIT lic github.com/pgbackrest/p...
github.com
GitHub - pgbackrest/pgbackrest: Reliable PostgreSQL Backup & Restore
Reliable PostgreSQL Backup & Restore. Contribute to pgbackrest/pgbackrest development by creating an account on GitHub.
040
Reposted by @qdrddr.bsky.social
PostgreSQL @postgresql.org · 16/07/2026
News: PostgreSQL 19 Beta 2 Released! www.postgresql.org/about/news/postg… #postgresql
031
qdrddr.bsky.social @qdrddr.bsky.social · 08/08/2026
@postgresql.org
010
qdrddr.bsky.social @qdrddr.bsky.social · 08/08/2026
#Firecrawl Anydoc fast #local Convertor: #Word, #PowerPoint, #Excel, #OpenDocument, #RTF, #EPUB, #CSV, and #PDF to clean #Markdown. Consume via #CLI or embed into your app: Built in #Rust, with #Node.js and #Python bindings. No LLM. For some complicated docs you still need an LLM #OpenSource MIT lic
github.com
GitHub - firecrawl/anydoc: Convert Word, PowerPoint, Excel, OpenDocument, RTF, EPUB, CSV, and PDF to clean Markdown. Built in Rust, with Node.js and Python bindings.
Convert Word, PowerPoint, Excel, OpenDocument, RTF, EPUB, CSV, and PDF to clean Markdown. Built in Rust, with Node.js and Python bindings. - firecrawl/anydoc
150
qdrddr.bsky.social @qdrddr.bsky.social · 07/08/2026
⭐️🔗 Docs [www.postgresql.org/docs/19/ddl-prop…...) 👉 Discuss with me in Discord: [linktr.ee/qdrddr](linktr.ee/qdrddr)
postgresql.org
5.15. Property Graphs
5.15. Property Graphs # A property graph is a way to represent database contents, as an alternative to the usual (in …
110
qdrddr.bsky.social @qdrddr.bsky.social · 07/08/2026
#PostgreSQL 19 brings #PropertyGraph capabilities to existing relational tables 🕸️🐘 Define: 🔹 Vertex tables 🔹 Edge tables from relationships Then query graphs with SQL + Cypher-like patterns. No separate #graphDB required. #PostgreSQL #SQL #GraphDB #OpenSource #LPG Repost 🔁 🔗 Link in first 💬⤵️
140
qdrddr.bsky.social @qdrddr.bsky.social · 07/08/2026
010
qdrddr.bsky.social @qdrddr.bsky.social · 06/08/2026
On-Device Maple-Preview 20B-A1B: Ternary-weight natively trained (not quantized) reasoning #LLM by Deepgrove. 131k context window. Mac mini M4, 218 tokens/s; iPhone 127 tokens/s. Consumes 7.7 GB of RAM. #AI #SLM #OpenSource MIT lic Model Weights & Blog deepgrove.ai/maple-preview
151
qdrddr.bsky.social @qdrddr.bsky.social · 03/08/2026
What about Apple’s own backdoor called “Suggested Places” in Maps app that executes arbitrary code on your device without your knowledge?
000
Reposted by @qdrddr.bsky.social
Cloudflare @cloudflare.social · 03/08/2026
gRPC support is now available in Cloudflare Workers. You can now route, transform, and serve gRPC services natively at the edge without maintaining dedicated proxy infrastructure or translation layers. Read the launch post: cfl.re/4fMsitT
blog.cloudflare.com
Cloudflare Workers and Containers now support inbound TCP connections and gRPC
Cloudflare Workers now support inbound TCP connections via Spectrum, allowing direct socket forwarding to Durable Objects and Containers. Developers can run full-duplex gRPC applications or leverage a...
2172
qdrddr.bsky.social @qdrddr.bsky.social · 03/08/2026
Fast local #PDF classification and data extraction, no #OCR. By firecrawl. Rust native, bindings for Python & TypeScript. 200 docs parsed in 0.470s. #OpenSource MIT lic. github.com/firecrawl/pd...
github.com
GitHub - firecrawl/pdf-inspector: Fast Rust library for PDF inspection, classification, and text extraction. Intelligently detects scanned vs text-based PDFs to enable smart routing decisions.
Fast Rust library for PDF inspection, classification, and text extraction. Intelligently detects scanned vs text-based PDFs to enable smart routing decisions. - firecrawl/pdf-inspector
050
qdrddr.bsky.social @qdrddr.bsky.social · 28/07/2026
Did you see any issues with Cloudflare MCP? It’s reasonable to use caching that can help.
010