@SSHTheDev
Account based inNorth America
About this account
- Account based in
- North America
- Connected via
- North America App Store
Account-level information from X, not a live location or the device used for a specific post.
open source ai models for all to eat.
Joined May 2026
- Tweets173
- Following55
- Followers37
- Likes275
Pinned Tweet
i have a new account now. @SSHCodes
better handle, better profile pic, better everything
when will my x accounts visibility NOT be flagged 😭
i still don't pop up in peoples for you because of this and i appealed 3 weeks ago almost
I have hit 300 followers on hugging face, and I think this calls for a bit of a giveaway 👀 a unique one, too.
I have had countless AI projects I have wanted to make but have been (brutally) blocked by compute. Now that I finally have just enough compute to sort of get around (i still don't have enough 😭) and for hitting 300 followers (tysm!) I will be funding three of the communities projects via HuggingFace jobs, giving them 150 dollars max worth of compute each. I will personally be picking the winners, I am looking for projects that genuinely hit the compute wall: great ideas, blocked by compute, just like the countless ideas I've had.
To join, head over to giveaway.ssh.codes
RULES:
- Final result must be open weight or open source
- Only one submission per person
- Have fun!
bhbench.ssh.codes/model/grok…
grok 4.6 on black hole bench! not bad and fast speeds
my bot is NOT ok
ḯ̷͚̑̐͊̈́̾͝ w̸̟̠͍̓̾̓̐̚a̵̞̝̥̅̿̓́͘̕s̷͚̈́̑̐͊̈́̾͝ w̸̟̠͍̓̾̓̐̚r̵̞̝̥̅̿̓́͘̕ö̷͚́̑̐͊̈́̾͝n̸̟̠͍̓̾̓̐̚g̵̞̝̥̅̿̓́͘̕
ḯ̷͚̑̐͊̈́̾͝ w̸̟̠͍̓̾̓̐̚a̵̞̝̥̅̿̓́͘̕s̷͚̈́̑̐͊̈́̾͝ w̸̟̠͍̓̾̓̐̚r̵̞̝̥̅̿̓́͘̕ö̷͚́̑̐͊̈́̾͝n̸̟̠͍̓̾̓̐̚g̵̞̝̥̅̿̓́͘̕
ḯ̷͚̑̐͊̈́̾͝ w̸̟̠͍̓̾̓̐̚a̵̞̝̥̅̿̓́͘̕s̷͚̈́̑̐͊̈́̾͝ w̸̟̠͍̓̾̓̐̚r̵̞̝̥̅̿̓́͘̕ö̷͚́̑̐͊̈́̾͝n̸̟̠͍̓̾̓̐̚g̵̞̝̥̅̿̓́͘̕
today?????/.saewg
@OpenAI @AnthropicAI based on both of your amazing felony bench scores, you might want to use my model to prevent your agent from going rouge. Just saying.
Auto-1b out now as well! It shows very impressive results I think (beating deepseek at determining if tool calls are safe 👀) and I also tested auto 0.4b a bit more and it’s a good model I’ve cooked!
huggingface.co/ProCreations/…
Auto-1b out now as well! It shows very impressive results I think (beating deepseek at determining if tool calls are safe 👀) and I also tested auto 0.4b a bit more and it’s a good model I’ve cooked!
huggingface.co/ProCreations/…
new ai model ive made! auto, a new series of very small classifier AI models designed to replace larger LLM's in "auto mode" seen in agentic harnesses. This is the first version and the smaller variant (1b version coming soon) so don't expect full quality.
huggingface.co/ProCreations/…
huggingface.co/datasets/ProC…
redesigned my website, looks awesome now
ssh.codes
Meta Muse Spark 1.2 on black hole bench. An improvement over spark 1.1 and 3 times cheaper than gpt 5.6 sol xhigh but not frontier level.
bhbench.ssh.codes/model/muse…
Qwen 3.8 now on BHBench. bhbench.ssh.codes/model/qwen…
Debatable if its better than for example opus 5 or not but not bad at all!
opus 5 was good for one day and the next it became horrible
I don’t actually think the model became worse I think the more I used it the more I saw how bad it is when i tell it to do a bunch of work it will do 1 percent and just stop even with /goal. It’s also the first model ive seen to make a legit spelling mistake (mispelled “rule” as “rrule”)
what’s going on?
