glub

6 ms·

61.
by glub 1mo ago
You may have more luck with GLM 5.3. New Kimi subscriptions are currently paused, so you have to join the waitlist. But even if you get in, usage limits are pretty bad there, just check reddit. GLM 5.3 is quite capable with "cyber&quot
62.
by glub 1mo ago
Ah yeah, new accounts get instantly blocked on TAC. Their backend has 2 failure states for `id_verification_status` - `failed` and `blocked`. Nationality bans get `failed`, new accounts (or rather, accounts with not enough good signals) get
63.
by glub 1mo ago
When they initially revoked TAC for a bunch of users due to a "technical error", the TAC verification flow opened a Persona iframe where you select the document country first. I was able to select it there (Georgia, in my case), w
64.
by glub 1mo ago
> Can ban you if you, in the "sole and absolute opinion" of Z.ai, have violated these broad terms. OpenAI revoked my Cyber verification, along with many others, asked to reverify (i.e. give my biometric information to Persona),
65.
by glub 1mo ago
Yes. I have no UI experience, and wanted a model that could produce something good without me telling it how anything should look like. My prompt was something like: "here's data I have, here's what matters to me, create HTML
66.
by glub 1mo ago
I don't know how anyone can actually use Luna max on ANY real workload. I've had Sol orchestrate a bunch of Luna agents, these agents were explicitly given small chunks of larger objectives and they still filled their entire conte
67.
by glub 1mo ago
You also make yourself extremely identifiable. It takes only one leak to know that all emails on mydomain.com across the web are delduca.
68.
by glub 1mo ago
Yeah, I don't really get the excitement. The pattern in which these aliases are generated might as well just end with private.icloud.com. Is breest-tas.7n@icloud.com really that hard to detect as hide my email?
69.
by glub 1mo ago
Anthropic's program doesn't apply to Fable. And OpenAI's one is silently limited by undisclosed list of approved nationalities.
70.
by glub 1mo ago
Despite having extensive security credentials, OpenAI revoked my cyber verification due to "technical reasons", with no ability to reinstate it, and without clear reasoning (it's just a generic error). But if I were to judge
71.
by glub 2mo ago
You're making a lot of assumptions, projecting, maybe? I do have dedicated machines and dedicated VMs for agents. RCE on a VM is still a RCE.
72.
by glub 2mo ago
Just a few years ago, people would lose their marbles if some software installed a background agent to send you notifications or something. Now we agreed that it's totally normal to have software that does remote code execution on our
73.
by glub 2mo ago
Tested muse spark 1.2 because it was rated so high on design arena, and I've missed a model that can do nice UI in the hands of an operator with no UI skills. It produced worse UI mockups than GPT and GPT models are already the bottom
74.
by glub 2mo ago
It's also starting to go beyond reasoning and it's becoming much more problematic. Reasoning is one thing, but codex, for example now encrypts agent-to-agent messages as well, and compaction. I've no idea what subagents are i
75.
by glub 2mo ago
GLM subscription is better than API, but significantly worse than Codex, even when used outside peak hours.
76.
by glub 2mo ago
If anything, it's going to be more expensive. Price/performance ratio isn't there yet for frontier open weight models. But regardless, you definitely should use a harness where switching models on the fly is easy. There'
77.
by glub 2mo ago
I've tested GLM 5.3 on the release day and Artificial Analysis is spot on. It's a really good model. But my main takeaway was something else. I've used closed weight models for long enough that I've forgotten how good it
78.
by glub 2mo ago
Yeah, for web apps, you can trick models by simply proxying it and pointing the models to that localhost. They then think they're not working on a live target. Have personally tested this with Opus and Sol and it works. Classifiers are
79.
by glub 2mo ago
Is this about the new project Blue and Red thing? Or just pre-existing Trusted Access for Cyber program? I've joined TAC, but still have to dance around it.
80.
by glub 2mo ago
Code owners - no. Ultra wealthy code owners with connections, and a few peasants with popular projects, for public image.
81.
by glub 2mo ago
No, you can give the model same prompt and it will give you a similar compaction result. On the backend, that's precisely what happens. There's nothing else going on in that encrypted blob, it's just summary of what model res
82.
by glub 2mo ago
The only "secret" there is a very basic instruction that the model receives, like "summarize current state and upcoming work" before compaction - same model that was just running your inference, with same cache, only ser
83.
by glub 2mo ago
I did this with Codex's recent encryption of compaction. Interestingly, I didn't have to drop to a dumber model, just a 2 sentence <developer> prompt auto-injected before and after compaction made all their models output the
84.
Third-party cyber evaluations involving OpenAI models(openai.com)
52 points by glub 2mo ago|7 comments
85.
by glub 8y ago
Location: Tbilisi, Georgia (UTC+4. But living at night, 9-5 PDT) Remote: Yes Willing to relocate: No Technologies: Python, Django, DRF, Elasticsearch (ELK stack in general - Kibana, APM, Logstash), Scrapy, Postgres, PostGIS, Celery, Spacy,
86.
by glub 8y ago
SEEKING WORK | REMOTE Role: Senior Backend Engineer / Architect with more than 6 years of experience. Architect and backend engineer of 10+ successful, large scale projects. Built distributed systems consisting of data crawling, pipeli