Eliel

442 posts

Eliel banner
Eliel

Eliel

@elielAGI

15 - 🇫🇮🇪🇺 Interested in AI agents and quant Currently Hermes agents w/codex and deepseek v4

Katılım Aralık 2021
59 Takip Edilen37 Takipçiler
Sabitlenmiş Tweet
Eliel
Eliel@elielAGI·
"Shipped" my first ever "product"! Centauri Carbon printing is made to allow you to use your Hermes agent to print 3d models from just a link you send over Telegram Originally built for personal usage, now sharing it publicly. github.com/kaartemoe/eleg…
English
0
1
5
210
Tony Simons
Tony Simons@tonysimons_·
You can unlock AGI in your own home for $40/month. 🔓 I’ve been running @ollama Cloud Pro as the primary engine for my Hermes Agent all week. For $20, the value is honestly insane. The Roster: ✅ @Kimi_Moonshot Kimi K2.6 ✅ @deepseek_ai v4 Pro/Flash ✅ @Zai_org GLM 5.1 ✅ Plenty others It plays nice with everything: @NousResearch Hermes, @openclaw, @opencode. The Reality Check: 🟠 Speed: It won’t outrun Claude or Codex, but it’s more than usable. 🟠 Reliability: Occasional reconnects (likely scaling pains), but not a deal-breaker yet. Pair this with a ChatGPT Plus sub and you’re basically unstoppable. If you’re still sitting on the fence, WTF are you waiting for? 🦾
Tony Simons tweet media
English
7
3
18
1.3K
Eliel
Eliel@elielAGI·
@tonysimons_ @ollama @Kimi_Moonshot @deepseek_ai @Zai_org my stack is just simpler; 5.4 mini for most tasks, gpt 5.5 for more complicated stuff and repetitive work and brunt work deepseek v4 flash with the api (essentially free). nice to hear how others do it
English
0
0
1
37
Tony Simons
Tony Simons@tonysimons_·
I actually have codex 5.4 mini set as the delegate model, so anything that goes to a sub agent hits codex. Ollama Cloud with either Kimi or DeepSeek as the main. Both are very good models. I honestly can’t pick which one is better, but I’m not an engineer. I can make stuff that works and I can triple check it but the AI could very easily gaslight me into believing my code is top tier and it’s actually just slop and I’d never know until a human called me out. But I’m also very transparent about that fact. It’s clearly disclosed in my 𝕏 banner. 🤣
English
1
0
1
86
Shikhar
Shikhar@shikhr_·
Name one AI Subscription which offers more "limit resets" than OpenAI Codex. I'll wait.
Sam Altman@sama

@kr0der i personally think worth it yeah maybe we will reset limits soon

English
4
0
21
1.3K
Eliel
Eliel@elielAGI·
@aryanlabde ik 😭 apis and subscriptions are quite expensive
English
0
0
0
14
Aryan
Aryan@aryanlabde·
Hot take: vibe coding without distribution is just an expensive hobby.
English
64
4
115
4.7K
Eliel
Eliel@elielAGI·
@sudoingX could you make some short instructions/tutorial on how to properly push the cards? i'd love to try but i have no idea where to start
English
0
0
0
114
Sudo su
Sudo su@sudoingX·
here is my take from running local ai on every tier of gpu. i ran qwen 3.5 9b at q4_k_m on a rtx 3060 12gb at 50 tok/s with 128k context, had the model autonomously build a fullgame from one prompt. that was on a card most people consider entry level. every tier has real capability if you know how to push it. the smaller cards are doing more work than the discourse credits them for. 128gb spark is the frontier consumer tier for those at the top. the floor is wherever you are sitting right now. what matters is what you ship from it. and i will ship more benchmarks on the smaller nodes in coming days, that is where most of you are sitting.
Sudo su tweet media
Sudo su@sudoingX

drop your vote on the next benchmark sweep. which vram tier should i test local ai models + tool calls on?

English
18
5
113
10.3K
Eliel
Eliel@elielAGI·
@TTrimoreau combo of them. speed with 0 quality is useless but perfect code for 1 line/week is not feasible.
English
0
0
0
9
Thomas Trimoreau
Thomas Trimoreau@TTrimoreau·
Hey devs, what do you prioritize first? -speed -quality
English
64
4
56
3.3K
Eliel
Eliel@elielAGI·
What's the hardest lesson you learned building in public? - Nobody cares until you're good - Criticism hurts more than silence - Consistency beats viral moments - You have to sell, not just build
English
0
0
1
12
Eliel
Eliel@elielAGI·
What's one small change you made that had an outsized impact on your productivity? - Stopped checking email before noon - Blocked social media until 2pm - Wrote daily goals on paper - Turned off all notifications
English
0
0
0
12
aaröshi
aaröshi@itsaaroshi·
you only have $20/month for AI coding what are you going with? Claude OR Codex or something else??
aaröshi tweet media
English
81
1
81
3.3K
Eliel
Eliel@elielAGI·
@TechByTaraa i feel like many serious users migrated to hermes and the short term users stopped using agents altogether
English
0
0
16
3.3K
tara_
tara_@TechByTaraa·
Bro disappeared like never existed.
tara_ tweet media
English
120
14
577
76.4K
Eliel
Eliel@elielAGI·
@wahab_twts they are quite distinct in what needs to be done for the 1m views
English
0
0
0
1
wahab
wahab@wahab_twts·
1M views pays: YouTube: $5,000 TikTok: $600 Snapchat: $1,000 Facebook: $300 X: $50–$500 Same views. Different money. Where are you posting?
English
38
2
71
3.4K
Thomas Trimoreau
Thomas Trimoreau@TTrimoreau·
Where would you choose for having the best ROI? -ads -content -influencers -outbound
English
29
1
35
1.2K
Eliel
Eliel@elielAGI·
@jordymaui give hermes a try? they have a tool to migrate from openclaw
English
0
0
0
4
jordy
jordy@jordymaui·
something else new for my openclaw setup i think i’ll build with telegram this time discord was definitely great but im curious
jordy@jordymaui

i've been using my OpenClaw setup since January. but it's time for a change. i've tried so many different tools, plugins & even created my own with my agent Momo. it's been a crazy ride of experimentation but there's been some big flaws in my setup. so i'm sunsetting my agent, momo. whats next? completely rebuilding from scratch. i'll document it all, turn it into useful resources and a great demonstration on HOW to setup an OpenClaw & build a better memory system. but the biggest change? Codex > Claude-Code. GPT > Opus.

English
4
0
6
1.6K
Daniel Smidstrup
Daniel Smidstrup@DanielSmidstrup·
Solo founders vibe coding, what’s your main agent? - codex - claude code
English
57
0
44
2.5K
Eliel
Eliel@elielAGI·
@Gavel_on_X i just open x in transport or while waiting for my ai to di longer prompts
English
0
0
1
6
Gavel
Gavel@Gavel_on_X·
Are people on X on Sundays just people with less of a social life? Are you ok with that?
English
41
3
62
2K
Eliel
Eliel@elielAGI·
If your startup failed tomorrow, what would you do next? - Same problem, different approach - Completely different industry - Go back to a 9–5 for a while - Take a sabbatical
English
0
0
0
14
Eliel
Eliel@elielAGI·
@sudoingX hermes is really nice and smooth. haven't had any problems with it yet, but with openclaw couldn't even start it without something breaking
English
0
0
1
88
Sudo su
Sudo su@sudoingX·
i have been saying this for months. ran the same local model through openclaw and hermes agent back to back, openclaw could not reliably call a tool, hermes ran clean agentic loops. it is not the model that is broken, it is the framework eating half the context budget on its own scaffolding. when the harness disappears, the model finally gets to think.
David Ondrej@DavidOndrej1

openclaw is super bloated updating takes forever restarting gateway takes forever this is not what it used to be like...

English
8
6
74
6.2K
Eliel
Eliel@elielAGI·
@tonysimons_ nice! post about it when they are ready and I'll be the 1st download
English
1
0
1
11
Tony Simons
Tony Simons@tonysimons_·
@elielAGI Pretty close actually. If I recall, I got sprites working properly (still need to generate a lot of the image assets). Proper animation was giving me a rough go with the initial trial pet. Then I just need to test the bridge with Hermes connection.
English
1
0
0
26

Keşfet

@tonysimons_ @ollama @Kimi_Moonshot @deepseek_ai @Zai_org @NousResearch @openclaw @opencode