ViktorKuz

Is a non-generative, retrieval-only architecture (excluding LLMs from the search steps) the optimal solution for building highly reliable and cost-effective personal memory systems?

ViktorKuz

7 mesi fadiscuss

More and more developers are switching to local LLMs - and the 1 reason is simple: security. Your data never leaves your machine. Zero risk of leaks. Meanwhile, we’ve seen dozens of high-profile incidents with cloud providers dumping private chats and prompts in the last 12–18 months alone. And you still have to pay premium for that “privilege”. At the same time, modern local models are basically on par with cloud ones. Qwen2.5-14B, Llama-3.1-70B Q4, or even 32B-class models now run on consumer hardware and deliver quality that’s within a few ELO points of GPT-4o-mini or Claude-3.5-Haiku — often beating them on specific tasks. This isn’t about “Chinese models suddenly winning”. This is about the future belonging to local optimization: quantization, speculative decoding, CPU offloading, MoE on a single GPU, etc. When you own the entire stack, you get speed + privacy + cost that no cloud provider can ever match. The tide has turned.

ViktorKuz

8 mesi fadiscuss

true the bottleneck was never "typing code", it was aligning business logic, constraints, and changing requirements. What AI does change is the cost structure: instead of one programmer spending a week, you can spawn 10 parallel agents and explore solutions in hours. The processes stay painful, but the iteration speed becomes insane and that changes who can build things.

ViktorKuz

8 mesi fadiscuss

Thanks for the kind words about VAC Memory System!

  Love what you're doing with ChatIndex - the hierarchical tree approach is really smart! Preserving all raw data
  while adding semantic navigation layers is an elegant solution. We're solving similar problems from different
  angles (you: lossless trees, me: gravitational ranking).

  Starred your repo! Looking forward to seeing benchmarks when you release them. Keep building!

Submissions

[untitled]

[untitled]

Hi, I'm Viktor.I wasn't a programmer. I didn't build apps. I didn't write code

80.1 % on LoCoMo Long-Term Memory Benchmark with a pure open-source RAG pipeline

Show HN: Change the model. Same output. The pipeline decides. VAC Memory System

My experience learning AI from scratch and why it changed how I see coding

[untitled]

[untitled]

Show HN: VAC Memory – 80.1% LoCoMo accuracy vs. Mem0's 68%

Show HN: MCA-Entity Coverage as a Memory Retrieval Gate

[untitled]

They Say AI Will Replace Programmers. I Think AI Will Mass-Produce Them Instead

I built an open-weights memory system that reaches 80.1% on the LoCoMo benchmark

80.1 % on LoCoMo Long-Term Memory Benchmark with a pure open-source RAG pipeline

Show HN: Memory System Hitting 80.1% Accuracy on LoCoMo (Built in 4.5 Months)

comments