#2ba — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #2ba, aggregated by home.social.
-
You don’t have to paste an API key into chat for it to end up in a prompt.
A coding agent can do it for you when it reads a config file or includes terminal output as context.
That’s why 2BA checks for secrets before prompts reach the model. When we detect an API key, we replace it with a dummy value. The model gets the context without the credential.
Security shouldn’t depend entirely on remembering to check everything an agent sends.
-
You don’t have to paste an API key into chat for it to end up in a prompt.
A coding agent can do it for you when it reads a config file or includes terminal output as context.
That’s why 2BA checks for secrets before prompts reach the model. When we detect an API key, we replace it with a dummy value. The model gets the context without the credential.
Security shouldn’t depend entirely on remembering to check everything an agent sends.
-
You don’t have to paste an API key into chat for it to end up in a prompt.
A coding agent can do it for you when it reads a config file or includes terminal output as context.
That’s why 2BA checks for secrets before prompts reach the model. When we detect an API key, we replace it with a dummy value. The model gets the context without the credential.
Security shouldn’t depend entirely on remembering to check everything an agent sends.
-
You don’t have to paste an API key into chat for it to end up in a prompt.
A coding agent can do it for you when it reads a config file or includes terminal output as context.
That’s why 2BA checks for secrets before prompts reach the model. When we detect an API key, we replace it with a dummy value. The model gets the context without the credential.
Security shouldn’t depend entirely on remembering to check everything an agent sends.
-
You don’t have to paste an API key into chat for it to end up in a prompt.
A coding agent can do it for you when it reads a config file or includes terminal output as context.
That’s why 2BA checks for secrets before prompts reach the model. When we detect an API key, we replace it with a dummy value. The model gets the context without the credential.
Security shouldn’t depend entirely on remembering to check everything an agent sends.
-
I asked 2BA Amber to build a Minecraft-style game set in space.
Then I thought: can it build the 2BA logo inside the game it just built?
So I let it try. :)
Here’s how that went.
-
I asked 2BA Amber to build a Minecraft-style game set in space.
Then I thought: can it build the 2BA logo inside the game it just built?
So I let it try. :)
Here’s how that went.
-
I asked 2BA Amber to build a Minecraft-style game set in space.
Then I thought: can it build the 2BA logo inside the game it just built?
So I let it try. :)
Here’s how that went.
-
I asked 2BA Amber to build a Minecraft-style game set in space.
Then I thought: can it build the 2BA logo inside the game it just built?
So I let it try. :)
Here’s how that went.
-
I asked 2BA Amber to build a Minecraft-style game set in space.
Then I thought: can it build the 2BA logo inside the game it just built?
So I let it try. :)
Here’s how that went.
-
When I say 2BA is “hosted in Europe”, what does that actually mean?
We run inference on GPUs we control, here in Europe. We’re not a proxy passing your prompts to another AI provider.
For me, this comes down to trust. If you’re sending us your code and your prompts, we need to know exactly where they go and what happens to them.
That’s why we want to control the whole pipeline, from prompt to response. It’s how we believe we can stand behind our promises of zero logging and European sovereignty.
Running the infrastructure ourselves takes more work. But that’s part of why I’m building 2BA.
-
When I say 2BA is “hosted in Europe”, what does that actually mean?
We run inference on GPUs we control, here in Europe. We’re not a proxy passing your prompts to another AI provider.
For me, this comes down to trust. If you’re sending us your code and your prompts, we need to know exactly where they go and what happens to them.
That’s why we want to control the whole pipeline, from prompt to response. It’s how we believe we can stand behind our promises of zero logging and European sovereignty.
Running the infrastructure ourselves takes more work. But that’s part of why I’m building 2BA.
-
When I say 2BA is “hosted in Europe”, what does that actually mean?
We run inference on GPUs we control, here in Europe. We’re not a proxy passing your prompts to another AI provider.
For me, this comes down to trust. If you’re sending us your code and your prompts, we need to know exactly where they go and what happens to them.
That’s why we want to control the whole pipeline, from prompt to response. It’s how we believe we can stand behind our promises of zero logging and European sovereignty.
Running the infrastructure ourselves takes more work. But that’s part of why I’m building 2BA.
-
When I say 2BA is “hosted in Europe”, what does that actually mean?
We run inference on GPUs we control, here in Europe. We’re not a proxy passing your prompts to another AI provider.
For me, this comes down to trust. If you’re sending us your code and your prompts, we need to know exactly where they go and what happens to them.
That’s why we want to control the whole pipeline, from prompt to response. It’s how we believe we can stand behind our promises of zero logging and European sovereignty.
Running the infrastructure ourselves takes more work. But that’s part of why I’m building 2BA.
-
When I say 2BA is “hosted in Europe”, what does that actually mean?
We run inference on GPUs we control, here in Europe. We’re not a proxy passing your prompts to another AI provider.
For me, this comes down to trust. If you’re sending us your code and your prompts, we need to know exactly where they go and what happens to them.
That’s why we want to control the whole pipeline, from prompt to response. It’s how we believe we can stand behind our promises of zero logging and European sovereignty.
Running the infrastructure ourselves takes more work. But that’s part of why I’m building 2BA.