kenize - Moozonian Search

news.ycombinator.com• Mar 4, 2026• 1 min read

Show HN: QLoRA fine-tuning in .zse INT4 format by ZSEReleased v1.4.0 of ZSE with QLoRA fine-tuning support for INT4 models.Verified benchmarks (H200 GPU, Qwen models):Model File Size VRAM (Inference) VRAM (+ Training) Speed 7B 5.57 GB 5.67 GB ~8 GB 37.2 tok/s 14B 9.95 GB 10.08 GB ~14 GB 20.8 tok/s 32B 19.23 GB 19.47 GB ~26 GB 10.9 tok/s 72B 41.21 GB 41.54 GB ~52 GB 6.3 tok/s What this means:Train 7B models on RTX 3070/4070 (8GB) Train 32B models on RTX 3090/4090 (24GB) Train 70B models on A100-40GB or 2x 3090Usage: from zse.format import load_zse_model from zse.training import LoRAConfig, add_lora_to_modelmodel, tokenizer, info = load_zse_model("model.zse") model = add_lora_to_model(model, LoRAConfig(rank=16, alpha=32))# Train normally, adapter is ~25MB save_lora_adapter(model, "my_adapter.safetensors") Trainable params: 0.2% of model (12M params for 7B)pip install zllm-zse[training]Code: github.com/zyora-ai/zse

realdeed.in• Feb 21, 2026• 1 min read

Show HN: RealDeed PropPass – Own Indian Real Estate Digitally from INR 10kHi YC / HNI’m Malhar, founder of RealDeed PropPass — and We’re building what we believe is India’s first structured tokenized real estate marketplace enabling users to digitally own a fractional piece of Indian property starting from just ₹10,000.The ProblemReal estate in India is:High entry barrier (₹50L–₹5Cr+ tickets)IlliquidPaper-heavyHard to verify titlesDifficult for young investors or NRIs to participateWhat We Built – PropPass by RealDeedPropPass allows investors to:Own fractional real estate digitallyInvest from ₹10,000Receive structured ownership rights via SPV modelAccess verified title documentationTrade fractional ownership (24/7 liquidity enabled model)Invest through escrow account mechanismsEach property is:Structured via an SPV (company/LLP model)Digitally represented as land-area-backed unitsLinked to verified property documentationManaged through compliant banking rails and escrow segregationWe are working with regulated banking partners for escrow management and clien

github.com• Jan 2, 2026• 1 min read

Show HN: Steganography in natural language using LLM logit-rank steeringI have had this idea rattling around my head for a while now. From the readme:> subtext-codec is a proof-of-concept codec that hides arbitrary binary data inside seemingly normal LLM-generated text. It steers a language model's next-token choices using the rank of each token in the model's logit distribution. With the same model, tokenizer, prefix, and parameters, the process is fully reversible -- enabling text that reads naturally while secretly encoding bytes.Basically, use the fact that LLMs learn a deterministic probability distribution over next-token generation to create seemingly innocuous ciphertext that is hard to detect.

bing.com• Oct 28, 2025• 1 min read

This New Index Could Be Crypto's Answer to the S&P 500A new index tracking the crypto sector was recently announced. It will track mostly crypto companies, along with a handful of leading cryptocurrencies. There will be a tokenized version of the index ...