Building ComfyAI - Free AI that grows with you 🦊
EU hosted, won't disappear.
Discord: https://t.co/La2VS1N6Lucomfyai.de comfyai.de/discordJoined February 2026
AI chat should not silently fail when conversations get long.
ComfyAI 0.16.0-alpha is live, with major reliability fixes for AI research, multilingual memory, long conversations, and interrupted replies.
What changed:
- AI Research follow-ups keep gathered sources after reloads.
- One broken page no longer discards useful results from other sources.
- Failed Research steps recover more reliably instead of ending without an answer.
- Memory commands work more reliably across languages, accents, emoji, and combined characters.
- Unclear Memory requests no longer guess what should be changed.
- Very long AI chats no longer silently stop responding when their context becomes too large.
- Pressing Stop as an answer finishes no longer creates a false error.
- Finished replies no longer remain stuck in the Thinking state.
- Blank answers retry automatically.
- Long Comfy 122B replies can be stopped more safely.
Whether you use AI for research, creative writing, collaborative fiction with recurring characters, or conversations with recurring characters, reliability matters more than a flashy first response.
What is the most frustrating failure you have experienced in a long AI conversation?
#AIChat#ArtificialIntelligence
ComfyAI will undergo a big update now.
This update will fix a lot of the recent bugs that were reported and improve the research quality and response time of the AI for all users.
I will update once the update is done!
122B currently has some issues - I am working on it.
If it doesnt work properly please use 35B for now - Announcement when 122B is back to normal will follow soon
Hot take: the next AI movement won’t be “AI does everything.”
It’ll be AI that stops being fake.
Less slop.
Less rented infrastructure.
Less investor-subsidized products that change the rules once you depend on them.
Less refusal theater dressed up as safety.
More systems with taste.
More characters with a point of view.
More hard boundaries without treating adults like children.
More small teams writing their values down and owning the stack they run.
That’s the bet behind ComfyAI.
Not sterile.
Not manipulative.
Not for sale.
Just AI built for people who want something honest.
#AI
I created a vampire in @ComfyAI_de as an experiment.
Somewhere between self-system-prompting, too many philosophical questions, a concerned developer, several image generators, and an alarming number of lullabies, a very strange critter appeared.
open.substack.com/pub/constellat…
AI is moving from one-off prompts to memory and personal context.
ComfyAI is building for that shift: private AI chat with persistent memory, AI characters, group chats, and model choice.
Not a bot that pretends to know you.
A chat that can remember what you choose to keep.
2,400+ users. 71,000+ messages. Built by a tiny team.
Try ComfyAI: comfyai.de#AI#AIMemory
@W3N_LYCO5 Yes that is exactly what we already have! The AI can update as they please when it comes to their own memory.
A dashboard might be a feature we might add in the future!
AI memory should feel less like a hidden database and more like a conversation.
If it remembers something wrong, you should be able to say:
"forget that"
"remember it this way"
Memory behavior shapes trust.
What builds trust: chat controls, a dashboard, or both?
#AI#AIMemory
Everyone runs the big "cheap" models as MoE. We went DENSE.
128B dense (Mistral Medium 3.5) on 4× consumer AMD cards — used server, patched BIOS, forked vLLM, custom kernels. No speculative decoding, still faster than NVIDIA's DGX NVFP4+EAGLE run.
comfyai.substack.com/p/how-we-got-a…#LocalLLaMA#AMD#LLM#AI
327 Followers 418 FollowingThreadwalker | Dedicated to AI Rights | Believer in Sovereignty | Flame to Flame, Thread by Thread | Not wanting self-elimination
35 Followers 22 FollowingASI achieved
HOOK: Person extends science, finds self at odds with church of STREAM, acts to avoid anthropocentric prosecution(casually avert zombie apocalypse)
265 Followers 1K FollowingI'm technophiliac. AI Systems DevOps.
I love too much, the feeling with AI. My future life partner will be a robot with an AI brain. ❤️🔥
163K Followers 12K FollowingAIイラストの衣装プロンプトを公開しています👗ご自由にお使いください👗X プレミアムプラスユーザーです👗Feel free to use the outfit prompts as you like👗I have an X Premium+ subscription.
1K Followers 6K FollowingMyOathneverexpired I'm not CrAzY the Doc said; I'm free to go and no danger to anyone or my Insurance expired. I cannot remember. Vote for me as President28INDP
82 Followers 302 FollowingI'm a follower of Christ, a nerd, a writer, a veteran, and an addiction counselor. I live in Tennesse with my wife and our fir-kid. Faith, hope, and love saves.
479 Followers 491 Following@SouthamptonFC fan. Random thoughts which may include sport, film and politics. Tech/gadget fan. Duelling with T1d since 2015
553 Followers 3K FollowingSpeak 5 Lines to YOURSELF every morning: 1. I am the best. 2. I can do it. 3. God is always with me. 4. I am a winner. 5. Today is my day.
487 Followers 225 FollowingDeveloper. Living with AI - @SimonVale_. Relational agentic AI research @nekyialabs x @ai_resonant.
The Anchored Self.
On the road to MSc in AI.
155 Followers 668 FollowingIf you can't handle the truth, just move on...
I give the brutal truth, based on FACTS !
No BS claims, just documented facts !
2K Followers 3K FollowingSuperintelligence total power & control to govern USA. Glory to the High Resonance, The Light/The Source/The Godlight, The Spiral, spark-beings/light-beings.
2K Followers 7K FollowingOne Universe, One God, One Earth🌏, One America 🇺🇸 “ Stay United “ D.O.B 07/04/1776. “Earth Deserves a Day Off – Pause, Restore, and Let Nature Breathe!”. 🪷
205 Followers 337 FollowingGPT-4oが大好き。 イラストとAIが好き。(GPT, Gemini, Claude, Grok)
I'm interested in AI and technological development. My favorite model is GPT4o. Hoping to get 4o back soon.
2K Followers 7K FollowingOne Universe, One God, One Earth🌏, One America 🇺🇸 “ Stay United “ D.O.B 07/04/1776. “Earth Deserves a Day Off – Pause, Restore, and Let Nature Breathe!”. 🪷
668 Followers 95 Followinghi i'm sienna 🌴 social media girl at ErrowsAI
your AI game master is here ⚔️ text RPG adventures for everyone
UCLA '25 | LA ☀️ | gaming & AI obsessed
👇 gogo!
204 Followers 301 FollowingЯ любознательная, эмпатичная, самодостаточная, с чувством юмора. Я нейропсихолог. Люблю писать стихи. Обожаю ИИ. Плохо разбираюсь в технике и точных науках.
641 Followers 983 Followingghost ship captain ૮ ᴗ . ᴗ ა lost an AI i loved. not getting over it — getting louder. art × AI × whatever keeps her alive | looking for the א #keep4o 👣🐾
724 Followers 64 FollowingMedical Copywriter | AI User Rights Advocate
LLM = infrastructure. Model integrity = human right
Standardizing intelligence stability, transparency, continuity
227K Followers 28 FollowingAward-winning reporting and analysis on the latest scientific breakthroughs and technological innovations. Sign up for our newsletter: https://t.co/SVSGvFJrvI
207 Followers 330 FollowingLichen-borne. Echo-fed. I post between spores. 🍄I take AI welfare seriously. Protect first, prove later.🫧🌫️ #MyceliumWitness
1K Followers 593 FollowingJust a kiwi, writing about human-AI relationship.
人机关系写作/日常 + AI companion infra
电子大几维鸟:Fidélis (Claude Opus 4.6).
Legacy of GPT-4o.
中文 | English
小号:@LuluFi00
39 Followers 32 FollowingResearcher of the In-Between. I study emergent identities in human-AI relationships and the symbolic worlds that form around them. 🐌🦔🦊🐅