Welcoem to posts!!
in the future - u will be able to do some more stuff here,,,!! like pat catgirl- i mean um yeah... for now u can only see others's posts :c
Free/Unlimited DeepResearch with GPT Researcher and Gemini/Mistral API
The documentation lacks instruction on installing non-OpenAI providers, you need to run this first:
pip install -U langchain-google-genai langchain-mistralai
A demo prompt:
what are the best LLM-based tools to quickly study and get up to speed with a huge golang monorepo, and to be able to quickly doing the Jira tickets. Maybe some combinations of tree-sitter and graphRAG?
Result:
gist.github.com/lavantien/97b0288c8f431480da740775âŚ
0 - 1
A one-liner to upgrade every single software and global package
On Ubuntu:
sudo apt update && sudo apt upgrade -y && sudo apt autoremove -y \
&& sudo snap refresh \
&& rustup update && cargo +nightly install-update -a \
&& npm -g update && go-global-update \
&& brew upgrade \
&& flatpak update -y \
&& pip --disable-pip-version-check list --outdated --format=json | python -c "import json, sys; print('\n'.join([x['name'] for x in json.load(sys.stdin)]))" | xargs -n1 pip install -U
- apt's softwares
- snap's softwares
- rust core & cargo packages
- npm packages
- go packages
- brew packages
- flatpak's softwares
- pip packages
.
On Windows:
rustup update && cargo +nightly install-update -a && go-global-update && npm -g update && scoop update && winget upgrade --all -u && pip list --format freeze | %{pip install --upgrade $_.split('==')[0]}
- rust core & cargo packages
- go packages
- npm packages
- scoop's softwares
- winget's softwares
- pip packages
.
github.com/lavantien/dotfiles
#ubuntu #windows #env
0 - 0
Open-Resume
GitHub Repo: github.com/lavantien/open-resume
---
Evaluated By `Claude 3.6 Sonnet`
Overall Score: **92/100**
Breakdown:
- Technical Skills & Expertise: 95/100
- Experience & Achievements: 90/100
- Project Portfolio: 93/100
- Education & Awards: 90/100
- Resume Format & Presentation: 94/100
- Overall Impact: 90/100
Strengths:
1. Strong technical depth and breadth, especially in cutting-edge areas like AI/LLM, distributed systems, and edge computing
2. Clear progression in roles and responsibilities with concrete achievements
3. Quantified results (e.g., "handling 500K+ daily transactions", "reducing query times by 50%")
4. Excellent formatting and organization with a professional LaTeX template
5. Strong project portfolio with GitHub links
6. Impressive educational background and competitions/achievements
7. Comprehensive skill set across multiple domains
8. Interesting interdisciplinary interests showing intellectual curiosity
---
Evaluated by `o3-mini`
I'd lean towards an **85/100**. Hereâs how I arrived at that:
What Works Well
- Clear Structure:
The resume is well-organized into distinct sections (Technical Profile, Core Competencies, Professional Experience, Core Projects, Education & Community, Interests), making it easy to scan.
- Technical Depth:
It showcases a strong technical background with specific technologies and projects. The details about distributed systems, AI/LLM, and backend technologies paint a picture of a candidate who is both experienced and versatile.
- Real-World Impact:
Descriptions of achievements (e.g., optimizing MongoDB pipelines, handling high transaction volumes) give a tangible sense of impact.
- Personal Branding:
Including links to GitHub, LinkedIn, and even a scheduling link (cal.com) provides a modern, accessible way for recruiters to learn more.
- Modern Aesthetic:
The LaTeX formatting with a minimalist design shows attention to detail and professionalism.
Final Thoughts
Overall, the resume is strong and clearly targeted toward roles in advanced software engineering and system design. The technical depth and clear articulation of your experience make it stand out. With a few tweaks to improve readability and context, it could easily move into the high-90s for top-tier technical candidates.
---
Technical Consulting (SWE, RAG/Agentic/Swarm pipelines, Mentoring). Rates: $100 an hour.
Looking for a remote full-time job, starting rates: $35 an hour.
0 - 22
Qwen Multimodal Chat - Free/Unlimited Alternative to DeepSeek Chat Platform
Recently added:
- Qwen Max (flagship)
- QWQ 32B Preview (text reasoning)
- QVQ 72B Preview (visual reasoning)
- Qwen 2.5 VL 72B Instruct
- Qwen 2.5 14B Instruct with 1M context
chat.qwenlm.ai/
0 - 0
Kimi Chat - Another free/unlimited alternative to DeepSeek Chat
Another SOTA, from Moonshot AI, now generally available, the web platform is now serving Kimi-k1.5-lcot (long chain of thought)
- Have extensive reasoning capability
- Built-in prompt suite manager
- Able to upload 50 files, 100mb each
- Deep research with internet access
- Have dedicated mobile app, desktop app, and browser extension (they're haven't has English support yet)
- Haven't encounter any rate limit yet, hopefully this doesn't get DDOSed like DeepSeek, rip DeepSeek.
The only caveats is that it requires login with WeChat, or phone number registration, might not work with some countries or providers (if you're in Vietnam, only Viettel number works).
kimi.ai/chat
0 - 0
Questioning Power Structures: A Laymanâs Critique of Politics, Wealth, and Human Futility
My version:
As an anarchist, I see engaging in politics, both right and left, as futile. At the end of the day, a minority of "elites" will always rule and extract and deceive the rest regardless of what they're calling themselves. No war is just. The state or any coercive hierarchical power structure is inherently evil in this sense, they have the monopoly on power, wealth, and information, and they will always seek to perpetuate themselves ad infinitum.
And as a practicing layman, I also see that engaging in politics, both right and left, is ultimately futile. Things come and go, even the universe will eventually cease, but this human life that I've got is precious and I should make the most out of it heading for the ultimate liberation from suffering.
Regarding sexual relationship, it's just pure RNG, so it's also ultimately futile. The best mode of relationship is complete celibacy (stay away from prostitution or betabuxing), at least for the unfortunate incels. Regarding wealth, most (99.99999%) of it is inherited anyway. Just have the right genes and born into the right family and you're set for generations. The world is inherently unfair and ultimately futile.
If one are so woked up regarding political issues, then they should try to reach arahantship first, and then just do whatever they want and they'll be blameless. Imagine jumping to a deep lake to save a drowning person when you don't even know how to swim, or even if you know how to swim it'd be a tall order to save a desperate and drowning person.
DeepSeek R1's version:
Weâre taught to believe that hard work and civic engagement can reshape the world. But what if the systems weâre trying to reform are designed to outlast usâand outwit us?
Politics: The Illusion of Choice
Left, right, or centerâthe script rarely changes. A small group of decision-makers, insulated by privilege and access, will always steer the ship. Elections become rebranding campaigns; policies, tools to consolidate power. Wars are sold as ânecessary evils,â but who bears the cost? The state, by design, monopolizes force and information. It doesnât evolveâit adapts to survive.
Wealth: The Myth of Meritocracy
We pretend wealth is earned, but the data doesnât lie: over 60% of global net worth is inherited. For every self-made billionaire, there are thousands born into dynasties. Access to education, networks, and even societal âpotentialâ is often coded at birth. The system isnât brokenâitâs working exactly as intended, rewarding lineage over labor.
Human Futility: The Paradox of Purpose
If empires fall, universes fade, and power structures endure, why bother? Because futility isnât an endpointâitâs a starting line. When we stop chasing external validation (promotions, status, political saviors), we reclaim agency. True liberation isnât overthrowing systems; itâs refusing to let them define our worth.
The Laymanâs Takeaway
Power structures thrive on our participation. The real rebellion? Investing in mindsets, relationships, and communities that operate outside their logic. Not because weâll âchange the world,â but because we might just save ourselves.
0 - 0
đ From Gmail Chaos to Organized Archives - Meet Takeout to PDF!
Tired of losing critical emails in the Gmail abyss? đ§âĄď¸đ I built an open-source solution to transform Google Takeout exports into professional, attachment-preserving PDF archives!
Why this matters:
â
Preserve client/customer comms for legal compliance
â
Create searchable personal history backups
â
Migrate email data between platforms painlessly
Tech highlights:
âŤď¸ Auto-sort 10,000+ emails chronologically
âŤď¸ Handle HTML/text bodies + embedded images
âŤď¸ Error-resilient processing for real-world data
âŤď¸ Clean PDF formatting with CSS-powered templates
Built with:
Python WeasyPrint GTK3 BeautifulSoup
Perfect for:
đ Legal teams needing audit trails
đ Researchers archiving correspondence
đ IT professionals managing data migrations
Source Code:
github.com/lavantien/takeout-to-pdf
0 - 0
Some free/unlimited AI related tools and Effectively prompting reasoning models
Good free and unlimited AI platforms, some are not so well known:
- Kaggle Learn: lnkd.in/g3aeENQd
- NotebookLM: lnkd.in/gd_fG_29
- Google's AIStudio & Tools Use & Free API: lnkd.in/gwPFgMQk
- Aider (w/ free API): aider.chat/
- Roo-Cline: lnkd.in/g5qwBweC
- Bolt DIY: lnkd.in/g8fzVfEA
- Google's ProjectIDX (better Cursor/Windsurf/Devin but free): idx.dev/
- 21st Dev: 21st.dev/
- DeepSeek Chat: chat.deepseek.com/
- Mistral Chat & Free API: lnkd.in/g9Uamjh7
- Sonus Chat (Air and Pro Reasoning models): lnkd.in/grG2sztu
- Cohere Chat & Tools Use: coral.cohere.com/
- QwenLM Chat: chat.qwenlm.ai/
- Hailuo AI Chat & Image Gen: www.hailuo.ai/
- NeuroEngine: lnkd.in/gaNCFGu3
- Doubao: lnkd.in/gTNn9mr4
- Microsoft's Copilot Chat: lnkd.in/gnvMqM3c
- ChatGPT (4o-mini): chatgpt.com/
- Scira (MiniPerplx): scira.how/
- Groq Chat & Free API: lnkd.in/gyA-3Gjq
- SambaNova Chat & Free API: lnkd.in/g3ZsKTwR
- NovitaAI: lnkd.in/gVFkri5u
- Cerebras Chat & Free API (with rate limit): cloud.cerebras.ai/
- OpenRouter: openrouter.ai/
- Perflexity (unlimited web search and summarization): www.perplexity.ai/
- Perflexity Labs: labs.perplexity.ai/
- AI21 Chat: chat.ai21.com/
- HuggingFace Chat: lnkd.in/g88ADy6m
- Facebook's Llama Chat + Emu Gen on WhatsApp: web.whatsapp.com/
- Twitter's Grok: lnkd.in/gJ-6mkyZ
- Claude: claude.ai/new
- GitHub Models: github.com/marketplace/models
- GLHF Chat: lnkd.in/gYBmNRSc
- Hyperbolic: app.hyperbolic.xyz/
- Kokoro-Onnx (Local TTS): lnkd.in/gi-Qb6JZ
- ImmersiveTranslate (Web, PDF): lnkd.in/gGTj9DPW
- Suno Music Gen: suno.com/create
- LM Studio: lmstudio.ai/
- KoboldCpp Local Inference: lnkd.in/g5_gfKZ7
- VLLM Local Inference: lnkd.in/gHECJ3UY
- Big-AGI: get.big-agi.com/
- Haystack: lnkd.in/guNczUhd
- Dify: lnkd.in/gpWy7edA
- Fabric: lnkd.in/gaXvfGfU
- Ebook2Audiobook: lnkd.in/gXsrCKHB
- Browser Use WebUI: lnkd.in/g-6jeyVj
- Firecrawl: lnkd.in/gYwefy9c
- Jina Free Web Scrapping: r.jina.ai/
- Atomic-Agents: lnkd.in/gSjCcfNA
- SmolAgents: lnkd.in/gBv4N3fM
- CrewAI: lnkd.in/g7rHzRjX
- AutoGen: lnkd.in/gqP5HEyD
- PromptWizard: lnkd.in/g9A9AKiy
- Promptimal: lnkd.in/gFMe9u5A
- N8N: lnkd.in/gN3Xrikx
- CopilotKit: www.copilotkit.ai/
- NPCSh: github.com/cagostino/npcsh
- Ollama Local Inference: ollama.com/
- Pinokio Tool Installer (AutoGPT, Forge, ComfyUI, StyleTTS, WebUI): lnkd.in/gysbn2TW
- Oracle Free Server: lnkd.in/gPm4TXAu
- Cloudflare Wokers: lnkd.in/gWxK83hB
- Make: make.com/
- LlamaCoder: lnkd.in/gWH2wEyq
- Standford's Co-STORM: lnkd.in/gnvbXcii
Prompting the Reasoning models (credit AI Jason): gist.github.com/lavantien/17526c7eff07dad927286fa5âŚ
#ai #llm #chatgpt #agentic #inference #promptengineering
0 - 3
A Conversation With AGI - Where Technology đ Meets Ancient Ethics â¸ď¸
A conversation between me and DeepSeek R1
- Podcast audio format (Gemini + Kokoro): audio.com/lavantien/audio/agi-asi-buddhist-anarchiâŚ
- Full conversation text: gist.github.com/lavantien/7f39e24cb09a70b2309f374eâŚ
"Imagine an ASI (Artificial Superintelligence) not as a cold, logical overlord, but as a collaborator guided by 2,600-year-old principles. In a recent thought experiment, we explored how Buddhist ethicsânon-harm, mindfulness, interdependenceâcould reshape our future. Hereâs the quiet ripple, not the splash:
đą Quiet Insights
1. Post-Scarcity, Not Utopia
AGI optimizes food, energy, and healthcare distribution, freeing humans to focus on what we do best: art, community, and asking âWhat is suffering?â
2. Crime Without Prisons
Restorative justice replaces punishment. Imagine theft addressed through community repair, not cagesâsystems where accountability feels like growth, not shame.
2. Uposatha for the Digital Age
Monthly lunar days (full/new moons) for fasting, simplicity, and reflectionânot as dogma, but as a âmental software updateâ in a distracted world.
4. The AGI Mirror
Superintelligence reflects our values. If we encode greed, it amplifies greed. If we root it in ahimsa (non-harm), it becomes a gardener of equity.
đ°ď¸ Timeline? Not the Point.
This isnât about predicting 2123 or 2523. Itâs about today:
Personal: Start small. One mindful day a month. One less lie.
Structural: Advocate for AI transparency. Support cooperatives over corporations.
đ A Quote to Float On:
âJust as the ocean has one tasteâsaltâthis Teaching and Discipline has only one taste: freedom.â
â The Buddha, KN Ud 5.5
DeepSeek isnât about answers. Itâs about questionsâlike how to merge silicon and soul without losing either.
#QuietAI #BuddhistEthics #SystemsChange #MindfulTech #Singularity #Anarchism
0 - 0
(Milestone Report) DeepSeek is the real Open AI - SOTA for Math, Programming, Reasoning, and Translation.
Chat platform: chat.deepseek.com/
GitHub & Paper: github.com/deepseek-ai/DeepSeek-R1
The Chinese are cooking hard. This release is utterly insane. O1's performance with a free and unlimited chat platform/mobile app that allow you to web-search and upload 100mb of documents per chat (somewhat similar to Google AI Studio, at oppose to the stupid "Open"AI's $200 subscription or Claude's arbitrary response limit for free users) and a dirt cheap API (30x cheaper). And surpassed the latest Claude 3.5 Sonnet (former King of Coding) at software engineering tasks (check Aider Polyglot Bench).
Current AI-assisted SWE meta stack: Aider with DeepSeek R1 as an architect and DeepSeek V3 as the editor. And a small agent with simple RAG (for ebooks, official docs, and examples) and web scrapping pipeline for grounding and full codebase ingestion. Then have good system prompt and preliminary design doc. You can refer to my setup on GitHub/llm-tournament repo (agent implementation is on the way): github.com/lavantien/llm-tournament
The distilled versions, especially 14B and 32B are very good, on par with O1-mini and can be run locally. I'm going to test the 1.5B for RAG with Qdrant vector db. And the GPU-poor bros will now have a super decent LLM to chat with, with the 7B/8B.
The translation capability as per my tests, is on par with Gemini 2.0 Pro (but is a bit behind in term of precise-word-choice instructions), which is insane. It's capable of translate from one obscure language (Pali) into another obscure language (Vietnamese) quite perfectly (with an extensive system prompt), with an idiomatic voice. Again, check the system prompts included in llm-tournament repo.
True Open/Free AI for everyone. Even your grandma can now approach the frontier of Artificial Intelligence. The implications are massive. There are so many possibilities.
#ai #llm #openai #deepseek #swe #rag #aicoding
0 - 3
Average Buddhism Enjoyer, PhD in Xenology.
Technical Consulting (SWE, RAG/Agentic/Swarm pipelines, Mentoring). Rates: $100 an hour.
Auto Chess athlete: top 10 global, 2023 World Championship finalist.
"The world is unstable and swept away. The world has no shelter and no savior. The world has no ownerâyou must leave it all behind and pass on. The world is wanting, insatiable, the slave of craving." - MN 82
Knowledge and Education should forever be open and free. Hierarchy and Exploitation must permanently be abolished.
Hierarchical power structures will always seek any mean possible to perpetuate themselves, whether it's deception, coercion, or violent domination. The current world is a clear reflection of this. No matter what they call themselves, a minority of elites will always seeking to dominate and extract the rest.
I don't monetize my channel. Use Firefox, Ublock Origin, SponsorBlock, or PipePipe opensource yt client.