Luke Celitan
@lukecelitan.bsky.social
Hello I am a tech lead and software engineer working in a 500 fortune company with more that 15+ years in software development and digital product creation
Senior Engineering pro tip: Always develop mobile apps on a real device and a bad / poor one. Most people don’t care about newest iPhones and flagship Samsungs, they just have their good old “shit” phone.
Buhahaha 🤣 … guys u seen this new hot project !!!! But seriously what I love about webpack is that you can chain it with TS properly, types fail = no build.
GitHub Trending JS/TS@github-trending-js.bsky.social · 2w ago🔥 Hot Repo！ 🔥 (100+ new stars) 📦 webpack / webpack ⭐ 65,985 (+105) 🗒 JavaScript A bundler for javascript and friends. Packs many modules into a few bundled assets. Code Splitting allows for loading parts of the application on demand. Through "loaders", modules can be CommonJs, AMD, E...
Small AI models + local infra = massive efficiency. 🚀 I've built a suite of tools to manage llama.cpp, vector DBs, and custom agents with integrations for GitHub, Google Workspace, and more. Check out the project here: https://github.com/MassivDash/ai_tools #AI #LLM #OpenSource
The AI hype cycle is intense. Is NVIDIA's stock holding up, or is the bubble inflating? What's the real story you're seeing? 🤔 #AI #NVIDIA #StockMarket
Tired of cloud AI costs & latency? I built a sovereign workstation agent in Rust/llama.cpp. Sub-second feedback. Zero token cost anxiety. Privacy: Nothing leaves your computer. Efficiency is the moat. #LocalAI #Rust #AgenticAI #SovereignWorkstation spaceout.pl/the-sovereig...
Solo effort vs. cooperative skill. Great software demands collaboration. That's where the real magic happens. ✨ #SoftwareDevelopment #Teamwork
Here’s the thing nobody tells you when they demo a shiny AI feature over a spreadsheet: the model reading your data and the model calculating your data are not the same skill, and one of them is a lie. Continue reading at spaceout.pl/ai-cant-coun...
Just hit a huge milestone: my first sale for S3 Manager! 🚀 I'm so excited to see people using this tool to manage their AWS S3, Backblaze B2, and Cloudflare R2 files. Thank you for the support! Check it out here: https://s3manager.spaceout.pl/ #AndroidDev #S3Manager #CloudStorage
Great software isn't built when an AI pats you on the back. It’s forged in the friction of human code reviews. We need collective ownership and critical peer feedback to build robust architecture. AI checks the code; humans build the system. 🚀
Everyone's chasing 2.8T parameter giants, but the real power lies in the niche. 🎯 A specialized small model + fine-tuning + smart tooling = the ultimate force multiplier. The future isn't just bigger models; it's smarter orchestration of specialized agents. 🛠️ #AI #LLM #SoftwareEngineering
The hard truth: Python is for the vibe, Rust is for the runtime. 🦀 Replace "Run -> Crash -> Debug" with a compiler-enforced feedback loop. Catch pointer errors early, stop wasting GPU cycles, and move from hope to safety. 🚀 #RustLang #AI #ML
Hard truth: Python is for the vibe, Rust is for the runtime. 🦀 Rust replaces "Run -> Crash -> Debug" with a compiler-enforced feedback loop. Catch pointer errors at compile-time to stop wasting GPU cycles. Move from hope to safety. 🚀 #RustLang #AI #ML
Dropped another fine tuned rust coder model, this time managed to squeeze 12B Gemma 4 model into my 5080 huggingface.co/MassivDash/G...
MassivDash/Gemma-4-RUST-CODER-12B · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
Architecture quality is invisible to the business, so it’s the first thing cut . Fitness functions fix that: stop arguing that coupling matters, let the build go red instead New post 🖐️ spaceout.pl/building-evo...
My rust small Gemma rust fine tuned model has been downloaded 900+ times in last 2 days … crazy stuff
Luke Celitan@lukecelitan.bsky.social · 4mo agoOn my way to build a small specialised rust codex model … fine tuned Gemma 4 today with unsloth studio huggingface.co/MassivDash/G...
On my way to build a small specialised rust codex model … fine tuned Gemma 4 today with unsloth studio huggingface.co/MassivDash/G...
MassivDash/Gemma-4-Rust-Coder · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
Recently I’ve been developing a s3 aws crud for android with heavy focus photo and video gallery, basically a s3 family photo vault app, I have decided to sell in Google play store, but I need few testers and early adopters, please dm me if you would be interested in becoming a tester !!!
Out of all vibe coding tricks, this one really makes a diff; run this prompt after whatever agent u use (cursor, antigravity, copilot) —> run a git diff and pretend you are a senior engineering manager that does a code review and does not like the implementation and then implement changes.
onyx.app/self-hosted-... … pretty good find from Reddit
Best Self-Hosted LLM Leaderboard 2026 | Open-Weight Model Rankings for Enterprise
The definitive self-hosted LLM leaderboard — ranking the best open-weight models for enterprise self-hosting across quality, speed, hardware requirements, and cost. Compare Llama, DeepSeek, Qwen, Mist...
onyx.app
Apple is no longer an exclusive company, run out off innovations and run into “cheap” market sector. Beginning of and end, gonna end up like Nokia in a couple of years.
Qwen3.5 9B is quite an impressive model, especially the vision features, I’ve been using it lately to create a catalog of my home library, it can read English and Polish titles … those small (Chinese) models are getting starting to exceed western models in my opinion
For decades, resilient software was built on the healthy friction of peer review. Today, AI coding assistants are creating a "solo dev army." But what happens when your pair programmer is a machine designed to agree with you? spaceout.pl/the-rise-of-...
Having your own portfolio / blog is always fun … you can show ur favourite star wars ships (at random) because why not ?
The AI adoption is indeed unprecedented … but seems at the cost of fundamental collaboration. Can really 1 man make a full movie with Ai ? can 1 man make a large scale software system ? Ai seem to think every out idea is great, but is it ?
I think people tend to generally over estimate the LLM capabilities, statistical machines that sequence words are not intelligence, chat gpt 5 cannot even count 5 different statuses on 100 entries properly …
Tried reading some of these moltbook entries, impossible… the site keeps on crashing. Some is about to get some massiv vercel bill for that s#tt. #moltbook
The AI hype around giant LLMs is fading—real-world AI is powered by smaller, smarter models. From fraud detection to real-time vision, here’s why production AI left the LLM “bubble” behind. Read more —>(spaceout.pl/the-great-corr…) #AI #TechTrends
