People familiar with the topic, how will models continue to get better? Post training it seems? Labs have already used up internet-scale data, so are there any limits to architecture improvements and post training or can we expect this trend to continue? ByteDance is training a 10T-parameter model. Here, GLM 5.3 outperforms models 3-4x its size of roughly 700B, so parameter count doesn’t seem to be a direct correlati…
GLM-5.3: Frontier coding with emergent cyber capabilities
441–450 of 626 posts
Re: GLM-5.3: Frontier coding with emergent cyber capabilities
#442Re: GLM-5.3: Frontier coding with emergent cyber capabilities
#443I'm not that up to date with the latest AI developments, but I noticed that this article seems to use "Cyber Capabilities" as a shorthand for the model's ability at cybersecurity tasks? Is that now an established expression, same as "crypto" now refers to cryptocurrencies rather that cryptography? Because "cybernetics" actually means something different (yeah, old man yelling at clouds, I know)...
Re: GLM-5.3: Frontier coding with emergent cyber capabilities
#444Earlier quoted context omitted.
I am not talking about GLM models being served through an Anthropic compatible API, that part is perfectly fine and I'm glad they support it! I am talking about ZCode, the program, being unable to delegate to other harnesses, like using Claude Code (or even OpenCode) within their UI, so that an Anthropic subscription can be used, because Anthropic don't let you use 3rd party harnesses directly. It's basically what Pa…
Ah sorry - I misunderstood. Thanks for explaining it. Have not heard of Paseo nor Kepler, and have never tried Zed. Yeah I too assumed if I'm to try use OpenAI subscription outside Codex, or Anthropic subscription outside Claude Code - I'd get my account banned it's agains their rules. So I have never looked how using the whole harness from outside looks like either (except for 'claude -p'). Interesting. BTW I see no…
Paseo had a really nice UI/UX, except sometimes sub-agents within OpenCode sessions would hang. Still, quite pleasant if you want something like the Codex or Claude Code desktop apps, but across various providers.
Kepler integrates with issue trackers like GitHub, you can just create a worktree from a ticket and let it churn, seemed like the second most polished option I tried, but there are obvious gaps - like moving cards manually, some missing UI options etc., which I'd chalk up to either the software just being that new or maybe being a little bit vibe-codey. Either way, one of the more promising options if you want something like Kanban board for agents.
Zed is mostly just a (really nice) text editor with some AI integrations, though it seems like they're also building a more agentic product as well - https://delta.dev/ haven't used that one much and am not in circumstances where I'd collaborate with people that closely, but there was a pretty cool podcast episode with the creators recently and it seems like it works pretty nicely for them! As an editor though, it succeeded where Fleet failed and has mostly replaced Visual Studio Code for me. Nothing against VSC, Zed just does most of the stuff I actually need out of the box.
Some of those tools interacting with Claude Code instead of trying to replace it is more or less the way to get Anthropic's models in other tools while still on a subscription (at least for now). How it works under the hood, go figure, there's ACP https://agentcommunicationprotocol.dev/introduction/welcome but also any number of hacky approaches.
To be fair, you can use Anthropic's models in many other harnesses directly, it's just that it then counts against API billing instead of your subscription, which ends up being way more expensive for individuals, but is kinda what you're supposed to do as a company.
Re: GLM-5.3: Frontier coding with emergent cyber capabilities
#445Earlier quoted context omitted.
I’m in the Claude code harness for everything boat too. What are the alternatives?
What the person above is suggesting: * https://pi.dev/ * https://omp.sh/ (no personal opinions of either, links might be useful) I think that OpenCode is nice, their CLI version is enjoyable and their desktop/web version is okay : * https://opencode.ai/ I also quite like driving OpenCode through something like Kepler / Paseo and tools like that (with those I can still use my Anthropic Condition by Claude Code being t…
Re: GLM-5.3: Frontier coding with emergent cyber capabilities
#446Feels like Fable's edge ended up just being long horizon task scaling, which post-training seems to achieve as seen here. Wonder what the next frontier is? Improvement in specialised tasks or computer use?
Anthropic needs to teach Opus how to speak English again, because Opus 5 seems to have forgotten. Utterly incoherent a lot of the time. They seem to be so busy scare-mongering and cooking up guardrails and watermarks that they haven't noticed that their models are getting weird.
Re: GLM-5.3: Frontier coding with emergent cyber capabilities
#447This is like when ‘crypto’ started meaning cryptocurrency.
Re: GLM-5.3: Frontier coding with emergent cyber capabilities
#448Earlier quoted context omitted.
ArtificialAnalysis puts out benchmarks for harnesses now as well, and OpenCode seems to be winning it. https://artificialanalysis.ai/agents/coding-agents#coding-ag... I only found this yesterday, and it inspired me to start testing out OpenCode.
Can OpenCode dispatch background subagents yet? I tried it a week ago and saw nothing. This is 99% of my workflow at this point.
Re: GLM-5.3: Frontier coding with emergent cyber capabilities
#449Re: GLM-5.3: Frontier coding with emergent cyber capabilities
#450I bought $18 GLM official subscription yesterday (5.2, but new model version was already leaking on some docs), set it up with Claude Code harness... and I’ve bumped to $80 plan almost immediately. It’s the first model that agreed on a proper security research (red team scenario), executed it seamlessly, including 0-days in WP plugins, RCE, 6.8 kernel exploit adaptation, etc - while playing against another GLM agent…