T1000@T1000_V2

AI enthusiast

Joined May 2025
39 Photos and videos
Ok you win. Opus is AGI
19
You know you can still build apps and websites with GPT 6 Pro in chat mode. One 6 Pro message is powerful enough to build you an entire app. If you keep getting blocked by Chatgpt when you ask, let me know. There are easy workarounds.
52
This self learning system benchmarks look promising.
Lakshya A Agrawal
@LakshyAAAgrawal
4 Sep 2025
Replying to @tobi
github.com/gepa-ai/gepa
1
2
256
Can Jev play online poker?
1
49
Why not task Astra with coming up with how to utilize Jev during every new large project/task. I can’t imagine the creative uses Astra could come up with.
1
43
It adds up ChatGPT Pro $100 Grok $30 Gemini $20 Midjourney $15 Hugging Face $10 OpenRouter $20ish Should I move to Chatgpt $200 and cancel everything else?
1
1
70
51
Unpopular opinion MD files are holding AI harnesses back. Using a database with multiple data types for each element in these files would allow dynamic updating and would allow for an interface to make changes like any other settings screen.
33
Pi using a sub agents extension running GLM 5.3 Flash is my new go to.
53
The only reason I still have a Gemini plan is they included youtube with 95% less ads via “youtube premium lite” . They found my weakness. I’ll never cancel.
1
78
Grok bots need cartoon voices
35
A $9 a month Comfy UI Pro subscription lets you generate videos with Minimax H3 for 40 minutes a day.
1
87
Grok Imagine
Made with AI
53
If you’re recommending an AI skill, you should show a before and after output to prove it works.
1
1
49
How is Qwen the only open source lab that can distill a coding model as powerful as Qwen 27b?
47
52
5.6 Luna Light can run Gauntlet Loops.
32
Instead of teaching an AI agent a skill through manually screen recording you doing it. What about just buying an online video course on Udemy and just feed all of it the agent?
43
Give Codex more capabilities at lower token costs utilizing N8N Extend capabilities with FFMPEG, Gotenberg PDF, Firecrawl PDF OCR or any useful open source tools you can install locally. Have Codex build your workflows in N8N & give your agent access to those workflows.
1
74
If Qwen 3.8 27B can be reduced to 20b - 22b parameters and match Qwen 3.6 27b quality, this would be an unlock for lower ram computers.
119
Load more