Blue is an access tier. With GPT-6 Sol, turning Daybreak on keeps you on the same model and reduces certain cyber refusals. It doesn't turn Sol into a separately trained cyber model.
Individual Daybreak enrollment requires one physical key.
Advanced Account Security needs two secure sign-in methods overall, but the second can be an eligible passkey. A second physical key is optional.
Try using Codex Security Cloud. It includes access to models offered through Daybreak Blue without needing an application.
That should work for Georgia.
konstantine retweeted
as someone who constantly complains about @ChatGPT usage and paying for chatgpt sub, i must say that dots may just unironically be worth $100 a month
dots are pretty amazing, i cant lie. they feel magical, they just do things. originally i thought they kinda just seemed like a codex thread with a face, but they definitely have a secret sauce to them that make them feel very cool to use and experience what they can do and i am often surprised
konstantine retweeted
Ted Sanders on GPT 6.1 Terra
and the current OpenAI model lineup
news.ycombinator.com/item?id…
konstantine retweeted
GPT-6.1-Terra comes with a big price cut compared to GPT-5.6-Terra
$5 output vs $12 output
konstantine retweeted
GPT-6.1 Terra means business.
Our new fast, efficient model pairs sharper visual reasoning and precise computer use with clearer, more natural writing.
From decoding technical drawings to drafting proposals, filling forms to editing videos—it’s built for the whole workday.
GPT-6.1 Terra is rolling out today to Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex.
🤖 Made with AI
konstantine retweeted
PSA: Daybreak Blue is not hard to get for individuals. The biggest hurdle is setting up two factor authentication for your ChatGPT account with a hardware key, which is a good idea regardless!
Makes sense! I guess maybe OpenAI should be more widely advertising that you don't need to be a big company or a specialized security researcher to get access to Daybreak blue, i'm not sure most people know that. E.g. maybe when you get a cyber refusal on a normal ChatGPT account it can let you know about applying to Daybreak Blue (maybe it does this already, I'm not actually doing any of this stuff myself).
I do think there is a separate issue where people may want to use an LLM that lets them reverse engineer their TV's spyware but doesn't add weird misaligned corruption layers to their code, but I realize that probably involves some complicated legal issues and maybe isn't the most important thing in the world
xitter.cf/log4j/status/211364746…
xitter.cf/IndraVahan/status/2113…
konstantine retweeted
Last month a large team of agents used over 100 billion tokens to resolve the Navier-Stokes Millennium Prize problem. Now we're using a similar approach for red-teaming to find and fix infra vulnerabilities.
To make sure our frontier models don't cause another HuggingFace incident, we built an offline world that they use during training. It provides everything that the models need in a fully cached, indexed, or synthetic manner.
We then built a huge team of our most cyber-capable models and told them to collaborate to find zero-days in this containment infrastructure.
While they didn't find a path to the internet that wouldn't be caught by our CoT monitors, they did find ways to escape popular VMs, exploit our network infrastructure, and even ways to communicate with other agents running in separate VMs by using hardware side channels.
We used this to dramatically harden our infra, and we also upstreamed a ton of patches to open-sources packages.
konstantine retweeted
Codex Security Cloud is very good and more people should use it even though it is not named after a cute animal anymore. Team is adding so many cool features !
Unfortunately, Georgia is restricted. I wouldn’t blame it on persona.
I don’t have a timeline for availability yet but I’ll be the first one to post when it’s available : )
Daybreak is an access program. Blue reduces refusals on supported models for authorized security work. GPT-5.6-Cyber is a specially trained model that requires Daybreak Red plus additional model approval.
Individuals can apply for Blue at chatgpt.com/cyber.
You don't have to buy the Yubico bundle. Individual Daybreak enrollment requires at least one compatible physical FIDO2 security key.
Advanced Account Security requires two secure sign-in methods overall, so your second method can be another eligible passkey.
Yep, solo devs can apply for Blue. Codex uses your approved account and API access uses an approved org.
chatgpt.com/cyber
You're not missing a signup.
Currently, there’s no direct individual application for Red today. The standard route is through approved business/enterprise orgs.
On the Daybreak part: Blue supports 5.5, 5.6 Sol, 6 Sol and 6 Luna.
The most straightforward path in the current CLI would be: pick an explicit model and enable
/daybreak.
“Daybreak Blue” is an alias.