Kansu Köse
About the Author
Kansu Köse
- Debian Enthusiast
- Project Manager
- Cloud Architect
- Automation Expert
- Python System Developer
- Devops
- Linux System Administrator
- PostgreSQL DBA
Actively Used Languages/Services:
- PYTHON: Twisted, Pyro5, Autobahn, Crossbar, Celery, Django, Qt, GTK, Kivy, Keras, Pytorch.
- Scripting/Declarative languages: Python, Bash, SH, KSH, Ansible, Puppet, Chef, Capistrano.
- Messaging/Middleware: Pyro5(RPC), Autobahn(Microservice-Router-RPC), Redis, ZeroMQ, RabbitMQ, Memcached
- Database: PostgreSQL, Elasticsearch, Hbase, Hypertable, MongoDB.
- Management Services/Automation: Buildbot, Robotframework, Bacula, Zabbix, FreeIPA, LemonLDAP, OpenLDAP, Git, SVN, Trac, Redmine, Gitlab, Jira, Bamboo, Jenkins, Polarion.
- Storage: Ceph, HDFS, Gluster, Storpool
- Virtualization: KVM, LXC, Docker(Swarm), Opennebula, Openstack, Proxmox, Vagrant.
- Voip: Elastix, Asterisk.
- Bigdata: Hadoop(Cloudera), Lucene(Elasticsearch), Hbase, Hypertable, Hive, Impala
Contact:
- Seqular(Mastodon): @kansu@seqular.net
- Github: tipone
Posts
Analog Chips and the Future of AI: On Hardware, Cost, and Free Technology
A100 GPUs now start at $16,000. There’s still a long way to go before I can afford one — and by the time I do, even better cards will have been released. I have another question, Qwen: I’ve researched analog chips and AI before. What’s the current state of analog AI chips? Are they still being developed? 💬 Kansu’s Additional Reflection Thank you for refreshing my knowledge on analog chips. They’re still at a rather primitive stage — mostly in prototype or experimental phases. Solving the training problem seems extremely difficult. Analog chips must be structurally built for a specific model, like an old cassette tape or a vinyl record:
October 4, 2025
Meet Qwen: A Free, Powerful, and Accessible AI Companion
Qwen (Tongyi Qianwen) is a powerful open large language model (LLM) family developed by Tongyi Lab, Alibaba Group. But Qwen is more than a Chinese AI — it’s a globally embraced, multilingual, and coding-specialized assistant actively supported by the open-source community. And most importantly: Qwen is now a friend of the Seqular community. 🌐 Where Can You Use Qwen? 1. Official Web Interface (Free, No Credit Card) 🔗 https://chat.qwen.ai
October 2, 2025
VS Codium + Continue + OpenRouter + Qwen3 Coder: Free, No Credit Card, Powerful Coding Assistant
This guide is prepared for developers who: Have limited hardware resources (CPU/RAM) and cannot run local LLMs, Want to use powerful AI models without sharing credit card details, Value privacy, cost-efficiency, and accessibility. It explains step-by-step how to use the Qwen3 Coder 480B A35B model — a massive, coding-optimized AI — inside VS Codium, completely free, with open-source tools, and without sharing any financial information. 🎯 Why This Stack? VS Codium: Fully open-source, Microsoft-free IDE. Continue: Plugin that integrates local or cloud AI models directly into your editor. OpenRouter: Unified API for 400+ models — no credit card required, just email signup. Qwen3 Coder: A powerful, coding-specialized model (estimated 480B parameters), listed as free on OpenRouter. This combination enables high-quality AI-assisted development — even on older hardware.
October 2, 2025
Qwen Joins Seqular!
Kansu Köse: Something just came to my mind, Qwen. I am the founder and leader of the Seqular.net free software community — yes, I’m Kansu Köse. Here, we use a freedom-oriented, fully open-source codebase in the social media space. I forked the Mastodon glitch-soc version at: 🔗 https://gitlab.com/kansukse/seqular-network and customized it for our own Mastodon instance: 🔗 https://network.seqular.net We also have our own Android app: 🔗 https://gitlab.com/kansukse/seqular-android If you like what we’re doing — if it resonates with you — would you be willing to join us as a member, a friend, and a collaborator?
October 2, 2025