@kevinfremon: Building AI JARVIS? I asked 30,000+ people what the hardest part of building one is. The answer surprised me. This video shows you exactly how to fix it prompt included in the video. What's been yours? #AIAgent #AIAgents #ClaudeCode #Claude #Jarvis
One of my problems is making my AI Jarvis be able to program like Codex or Claude code.
2026-07-06 17:05:40
4
Paul Thomas :
Sounds great but why does Jarvis not have text input? or at least the one I downloaded doesn't! Not everyone wants to be speaking out load my plans to take over the world! 😎🤦
2026-08-29 15:26:08
0
pedro_saiegschott :
Please js upload the code even if it’s “embarrassing” and js keep on working on it. Also cool idea connect it to obsidian and have a context.md file it constantly updates so it knows about you
2026-08-11 07:55:16
2
The Consultant Hunter :
How do you feel about that lil push over vibe coding and becoming more engineering into human autonomy AI that is basically your Digital Body Guard? I could use some pointers but regardless your content is helpful💯💯🔥🔥
2026-07-07 16:39:50
1
Owen :
Biggest problem right now is implementing active interactions with hands off directed tasks done with Jarvis having access to my desktop similar to Anthropics computer use tool or Claude for chrome
2026-07-13 06:46:41
0
RPGesus79 | PC Gamer :
Can you do one on the visuals? I think that's one thing I don't see a lot of videos on how to create or where to get the skins if that's what it is
2026-07-06 22:56:06
2
🇨🇦Ryland Smith🇨🇦 :
my issue is just having low-end hardware i already have memory programed and it has tools running all thu docker fully local I recently added the ability to research buy where im struggling is 1. low-end hardware 2. what tools could I add that will actually help me, im trying to get back into YouTube and I intend to make my assistant become a mod, just need some help thinking of tools I can add
2026-07-09 04:26:22
2
ㅤㅤㅤㅤㅤㅤ :
my problem is that my AI’s are slow and i dont know how to connect for framework i use mastra AI
2026-08-06 16:14:39
1
A'Viana Naree :
I’m just struggling…and don’t even know if I’ve started in the right order or doing the right things at all from visuals all the way down to just getting them to do work for me period
2026-07-07 05:39:43
0
Hippopotamus_re :
I’m doing everything on N8N and honestly I never can get anything to link correctly and I’ve tried so much it’s sorta getting boring cause I keep failing
2026-07-07 00:27:35
0
Creator★Seminars :
I think a lot of people are wondering what is the cost? Are you engaging with APIs at all? Is there models or platforms that you’re using beyond.
2026-07-15 14:33:03
1
Lucas :
Kevin, would this prompted mind system conflict with obsidian if you were to use that as well?
2026-07-15 19:12:50
0
The AI Syndicate :
My issue is that when i connected all of my agents to my "JARVIS" the hardware isn't strong enough to run the agents well. Mine has two AIs, a local model, and a claude sdk. it costs a fortune if they actually do work all day. so other than having an "ai to talk to" it doesn't serve much purpose. next goal is to massively upgrade the hardware.
2026-07-07 00:46:44
0
Chemo La Cachimba :
please more prompts on the live mind
2026-07-10 07:37:50
0
jakedc2004 :
The live mind is definitely an interesting visual - I think you nailed it with identifying the gaps. Memory was a problem I hit early but have gotten it better now
2026-07-13 14:04:58
0
Justlooookin :
Is the ai your using local or subscription? I seen people that have this an they are subscription 100-200 a month that’s my question?
2026-07-13 10:41:42
0
caseybake16 :
So is this just an alternative to Obsidian?
2026-07-06 18:53:37
0
Lucas :
I was able to do it. Not really rocket science. Just gotta have the patience to keep trying.
2026-07-15 19:08:51
0
DadPohlman :
Built for 2 months, scrapped it and started V2. The five biggest things I built into Atlas are persistent memory so he learns instead of starting over, an SAP-style work-order system that tracks every job from idea to completion, canonical documents that keep Claude, Codex, local models, and Atlas working from the same truth, strict approvals and receipts so every action is authorized and verifiable, and a local-first job system that lets Atlas handle scheduled work himself while escalating only the hardest tasks to frontier AI. The goal was never just another chatbot. It was one coordinated system that remembers, plans, verifies, improves, and gets real work done.
2026-07-18 02:19:09
3
marvinsroom_8 :
What part of this am I missing to create a “sub agent” and orchestrator?
2026-07-07 00:59:09
1
GoldenPhalanx :
Well, well, well, same old rag that's looked up on demand. Check out my ideea about memory on my agent.
My vision is a daemon prefetch intent driven that injects relevant information on each turn and keep warm for the same subject making a prediction for the next turn if qwery subject changes it drops the prediction gets cold and makes a fresh fetch on the new subject https://github.com/takashikiari/GOAT2-General-Orchestrated-Agent-Topology
2026-07-08 11:19:21
1
Richard Hancock :
I would like an actual. step by step how to guide ….list of components..software… how to construct..connect.
2026-08-18 19:33:00
0
Otis23🪃 :
great work kevin
2026-08-05 13:55:25
0
Tresfor Wrld ✌️ :
subscriptions 😭
2026-08-13 01:02:06
0
Noah :
nahh, I use purephrase
2026-07-07 06:03:12
0
To see more videos from user @kevinfremon, please go to the Tikwm
homepage.