DeltaSqueezer

• AI assisted music creation

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 6 comments

• JetBrains open-sources Mellum2 - anyone tried these?

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 10 comments

• A moment of thanks for DeepSeek

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 22 comments

• New DeepSWE benchmark finds Claude Opus cheats

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 92 comments

• Fed up with vibe coders, dev sneaks data-nuking prompt injection into their code

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 135 comments

• The power of structured workflows and small local models

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 45 comments

• got my first "rm -rf /" today

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 143 comments

• CSP and z3 solver

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 2 comments

• Small local model for questions on German grammar

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 24 comments

• Prompt injection testing

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 5 comments

• AI startup Cohere valued at $6.8 billion in latest fundraising, hires Meta exec

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 24 comments

• Could PC x64 instruction extensions relieve hardware shortage?

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 19 comments

• What questions have you asked reasoning models to solve that you couldn't get done with non-reasoning models?

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 34 comments

• Recommendation for open source LLM proxy software

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 11 comments

• Home-rolled loop agent is surprisingly effective

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 42 comments

• Warning: do not write your own AI agent if you don't want to get sucked into a blackhole

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 6 comments

• Warning, if you are on the spectrum, do not write your own AI coding agent - it is too addictive

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 0 comments

• AI Coding Tabs vs Spaces

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 2 comments

• How do you stop codebase from degenerating into an un-maintainable AI-slop mess?

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 87 comments

• What open weight model for diagram generation

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 8 comments

• How do you monitor what an agent is doing?

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 1 comments

• Missing a Qwen3.5 model between the 9B and the 27B?

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 66 comments

• Qwen 3 30B-A3B on P40

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 26 comments

• ARC-AGI-3 is a fun game

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 5 comments

• TurboQuant: Redefining AI efficiency with extreme compression

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 0 comments

• Running Qwen3.5 in vLLM with MTP

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 8 comments

• Kokoro #1 on TTS leaderboard

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 89 comments

• Ruthless prompting efficiency

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 4 comments

• vLLM profiling of prompts

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 1 comments

• Finally, a real-time low-latency voice chat model

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 463 comments

• Alibaba Releases OpenSandbox to Provide Software Developers with a Unified, Secure, and Scalable API for Autonomous AI Agent Execution

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 2 comments

• Reverted from Qwen3.5 27B back to Qwen3 8B

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 39 comments

• Llama 3.2 1B surprisingly good

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 55 comments

• Is there a way to cleanly terminate a running inference job/slot with llama.cpp?

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 2 comments

• Local Text To Speech

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 41 comments

• Weird Qwen3.5 27B 'rabbit hole' failure mode

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 12 comments

• OK, now we're at 1T parameter models, what's the 3090 equivalent way to run them locally?

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 62 comments

• Hardware ASIC 17k tok/s

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 6 comments

• Could High Bandwidth Flash be Local Inference's saviour?

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 29 comments

• Re-sizeable BAR (REBAR) impact on performance

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 16 comments

• China’s AI is quietly making big inroads in Silicon Valley | Technology

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 13 comments

• GitHub - gruai/koifish: A c++ framework on efficient training & fine-tuning LLMs

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 5 comments

• Moving beyond vibe-coding

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 10 comments

• OCR tools for really very bad handwriting!

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 80 comments

• Transparent LLM logging proxy

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 5 comments

• GLM 4.7 on 8x3090

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 37 comments

• Cheap bifurcation

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 20 comments

• Nvidia CMP 170HX for LLMs?

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 21 comments

• GLM released 4.6V including the apparent successor to Air. But I'm most interested to test the 9B Flash version

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 22 comments

• llama.cpp not using kv cache effectively?

  Posted by DeltaSqueezer@reddit | LocalLLaMA | View on Reddit | 17 comments