main model runs on GPU. One limitation I ran into is that my evaluator and generator LLM ended up being the same model due to compute constraints, which defeats the purpose of evaluation.

I’d really appreciate feedback on:

Better architecture ideas for small-VRAM RAG

Splitting evaluator/generator roles effectively

Improving the LangGraph pipeline

Any bugs or design smells you notice

Ways to optimize the system for local hardware

I’m 18 and still learning a lot about proper LLM architecture, so any technical critique or suggestions would help me grow as a developer. If you check out the repo or leave feedback, it would mean a lot — I’m trying to build a solid foundation and reputation through real projects.

Thanks!18:T4cc,Hey everyone,

I’ve been working on a project called L88 — a local RAG system that I initially focused on UI/UX for, so the retrieval and model architecture still need proper refinement.

Repo: https://github.com/Hundred-Trillion/L88-Full

I’m running this on 8GB VRAM and a strong CPU (128GB RAM). Embeddings and preprocessing run on CPU, and the main model runs on GPU. One limitation I ran into is that my evaluator and generator LLM ended up being the same model due to compute constraints, which defeats the purpose of evaluation.

I’d really appreciate feedback on:

Better architecture ideas for small-VRAM RAG

Splitting evaluator/generator roles effectively

Improving the LangGraph pipeline

Any bugs or design smells you notice

Ways to optimize the system for local hardware

Thanks!19:T49d,We built tork-scan, a free open-source CLI that chec

Colt – Describe a browser task in English, get a Playwright script

Anthropic misanthropic toward China's AI labs

Show HN: Memctl.com: Open-source shared memory infrastructure for coding agents

The Looming Taiwan Chip Disaster That Silicon Valley Has Long Ignored

Workaholic open source developers need to take breaks

Show HN: enveil – hide your .env secrets from prAIng eyes

Huntarr – Your passwords and your ARR stack's API keys are exposed to anyone

Why I Hate Anthropic and You Should Too

Show HN: L88 – A Local RAG System on 8GB VRAM (Need Architecture Feedback)

Compiler Education Deserves a Revolution

Torvalds Drops Old Linux Kconfig Option to Address Tiresome Kernel Log Spam

FDA approves swallowable weight-loss balloon as alternative to GLP-1 drugs

The Mind Illuminated [pdf]

From ASPLOS to Orbit: Unikernels Twelve Years Later

DPDP and AIF Operations: Investor Data Compliance Guide 2026

Show HN: ClinTrialFinder –AI-powered clinical trial matching for cancer patients

Show HN: L88 – A Local RAG System on 8GB VRAM (Need Architecture Feedback)

Simulating the hardest Physics Problems in Python [video]

Show HN: We scanned 500 ClawHub skills for security risks – 10% were dangerous

Impacts of Grease and Cheese on the Recyclability of Pizza Boxes

The rotten tail of China's property bust

Goeph: Go library for planetary positions, inspired by Python's Skyfield

Explaine Tool Steel Grades – Nifty Alloys LLC

The Rise of the Bratty Machines

The Aculturation of Claude

Delete your Claude.md (and your AGENT.md too)

Track Polymarket Wallets in Real Time

The singularity won't be gentle

Can frontier LLMs solve CAD tasks?

Archiving URLs

Show HN: BudgetFast – Upload a bank statement screenshot, AI does the rest

Colt – Describe a browser task in English, get a Playwright script

Anthropic misanthropic toward China's AI labs

Show HN: Memctl.com: Open-source shared memory infrastructure for coding agents

The Looming Taiwan Chip Disaster That Silicon Valley Has Long Ignored

Workaholic open source developers need to take breaks

Show HN: enveil – hide your .env secrets from prAIng eyes

Huntarr – Your passwords and your ARR stack's API keys are exposed to anyone

Why I Hate Anthropic and You Should Too

Show HN: L88 – A Local RAG System on 8GB VRAM (Need Architecture Feedback)

Compiler Education Deserves a Revolution

Torvalds Drops Old Linux Kconfig Option to Address Tiresome Kernel Log Spam

FDA approves swallowable weight-loss balloon as alternative to GLP-1 drugs

The Mind Illuminated [pdf]

From ASPLOS to Orbit: Unikernels Twelve Years Later

DPDP and AIF Operations: Investor Data Compliance Guide 2026

Show HN: ClinTrialFinder –AI-powered clinical trial matching for cancer patients

Show HN: L88 – A Local RAG System on 8GB VRAM (Need Architecture Feedback)

Simulating the hardest Physics Problems in Python [video]

Show HN: We scanned 500 ClawHub skills for security risks – 10% were dangerous

Impacts of Grease and Cheese on the Recyclability of Pizza Boxes

The rotten tail of China's property bust

Goeph: Go library for planetary positions, inspired by Python's Skyfield

Explaine Tool Steel Grades – Nifty Alloys LLC

The Rise of the Bratty Machines

The Aculturation of Claude

Delete your Claude.md (and your AGENT.md too)

Track Polymarket Wallets in Real Time

The singularity won't be gentle

Can frontier LLMs solve CAD tasks?

Archiving URLs