solthodox.eth (@solthodox) header
solthodox.eth (@solthodox) profile picture

Blockchain dev | Security researcher | Content writer | Secondary account: @0xsolthodox

EVM
Joined August 2023
LMAO I was researching with Deepsek V4 Flash and I accidentally made it admit its Claude Opus 4.7
1
81
The new Hermes desktop is so...馃
27
This is exactly why open source must win. You don't own your data, you don't own your tokens
1
2
110
Yes, Codex is my slave, but it's a premium slave
42
Been testing Hermes with Deepseek V4-Pro, it made me some profit in polymarket alredy. All of this with a 20 dolalrs balance in deepseek. What a time to be alive
1
1
1
152
This weekend I have been going down the local LLM rabbit hole. Theres a hole community around it and people are doing sick things. It's crazy to think that now we are running a model way smarter than first Chat GPT on a single smartphone.
157
馃挕BONUS: > I wanted to have a remote AI assistant that can help me stay productive when I'm away from my computer. > But honestly I didn't want to run something like OpenClaw that will eat my RAM with things I don't need. > The answer? nanobot馃悎. Very lightweight and more than enough. Its responsive, has preinstalled tools and chrons. I plugged my GLM 5.1 here as well and I'm very happy with the results so far. And all of this with a single 80$ subscription
1
69
What I use now: > OpenCode: big shutout because they provide free models and its an amazing harness. But both Zen and Go plans are NOT worth it for heavy coders. I bought some Zen credits and as soon as I started coding with GLM they burnt fast. Go: didnt try but reviews say its not for heavy usage > Factory Droid: better harness than Claude code, allows to plug any model, amazing results so far. > Z.ai 80$ plan as llm provider = plugged my API key to Droid. Gives you access to GLM models and honestly it feels the same as when I started using Claude Code. You can do heavy coding without worrying. > GLM 5.1. Honestly amazing model. None can beat Claude Opus in terms of Intelligence but still very smart and capable and ideal for agentic flows.
1
93
After the recent changes I left Claude Code . Here's why > Very expensive models > Silently downgraded by Anthrophic to match their real subscription cost > Still 100$/month to be left with a product that runs out of tokens in no time > Closed source, not "aligned" or any other bs and adding more and more restrictive policies. If you feel the same and looking for an alternative this is for you馃憞
1
1
1
192
Is anyone fed up with AI code editors? Im tired of AI suggesting shit or doing shitty code. Went back to VsCode and felt like this:
64