serrated

express intent through constraint

[serrated @SerratedLiminal]

@PetuniaWil65888 Well preserved skeleton

[serrated @SerratedLiminal]

@word0taku Procreate

[serrated @SerratedLiminal]

@Oggle2016 @vaxryy @CenturiiC She doesn't have enough clothes to cover herself properly.. M

[serrated @SerratedLiminal]

@thsottiaux @theo What's the sweet spot for context size? Cost and compaction frequency wise.

[serrated @SerratedLiminal]

@ammolab1 How are you gonna aim?

[serrated @SerratedLiminal]

@maria_rcks @thsottiaux They did limit reset too

[serrated @SerratedLiminal]

@usr_bin_roygbiv Luna is cheaper even though it overthink (more reasoning hoops) and consume more tokens. Sol is decisive but it follows instructions more faithfully.

Terra is a good middle ground.

[Tibo @thsottiaux]

Morning. The last 48 hours of Codex and ChatGPT Work have been intense! Three important updates:

- Temporarily removing the 5 hour usage limit restriction for all Plus, Business and Pro plans
- Rolling out changes that will make GPT 5.6 Sol more efficient across the board and that will be reflected in less usage being used so that it can take you further. Exact impact to be quantified and shared
- We hit 6M active users, and are landing a usage reset in the next hour

Go do things

[serrated @SerratedLiminal]

@nicdunz I observed this earlier on Claude.

[serrated @SerratedLiminal]

apparently anthropic is using gVisor to run a sandbox, real file commands work, but kernel + dmesg are virtualized. syscalls are intercepted.

I crashed the container twice by running python stress test.

And found something even more interesting on my third run. I asked it assume 7 cpu even though nproc was 2.

[serrated @SerratedLiminal]

@nicdunz Not yet, but I think they may migrate to it eventually.

Container brings codex like features on chatgpt web, which would greatly benefit corporate users.

[serrated @SerratedLiminal]

I think ChatGPT's work is closer to a container-based persistent execution env in Claude. But Claude has actively optimized for it, whereas Codex did not.

That could be a possible reason for ChatGPT and work to coexist, at least for now.

[serrated @SerratedLiminal]

Perfection doesn’t need love because it has no lack, no desire, and no serrated edges.

Loving is seeing, seeing the serrated edges and still loving.

[serrated @SerratedLiminal]

@notiyda @xadisingh Is thinking or waiting for results?

[serrated @SerratedLiminal]

@DrMacEachern @nicdunz I am getting the same issue. I close and reopen codex and a few percentage drops.

[serrated @SerratedLiminal]

@Azurei_7 @markgurman Probably using android

[serrated @SerratedLiminal]

It's strange that 5.6 sol behaves in same way as grok when processing ambiguity. Both follow instructions more faithfully.

[serrated @SerratedLiminal]

I think ChatGPT's work is closer to a container-based persistent execution env in Claude. But Claude has actively optimized for it, whereas Codex did not.

That could be a possible reason for ChatGPT and work to coexist, at least for now.

[serrated @SerratedLiminal]

@LinQi4ever @sama I'm glad they nerfed it

[Artificial Analysis @ArtificialAnlys]

GPT-5.6 Sol and Luna are ahead of Terra at every point on the Intelligence vs Cost per Task chart. GPT-5.6 Luna stands out as a particularly cost efficient model

Charting the Artificial Analysis Intelligence Index shows the trade-off between intelligence and Cost per Intelligence Index Task. Across reasoning efforts, each GPT-5.6 model pushes past GPT-5.5 on the Pareto frontier (excluding non-reasoning).

However, Luna and Sol are always ahead of Terra. This means for any Terra effort level, there is a Luna or Sol effort level that is more intelligent at no extra cost, or as intelligent at lower cost.

[serrated @SerratedLiminal]

At this point, only Luna looks suitable for me. It's cheaper than 5.5 medium and scores higher.

[Sebastian Raschka @rasbt]

For agentic coding, one can say:

- Unless you need Terra Ultra perf, it's always better to use a Luna model with higher effort setting (same or better performance but cheaper).

- Forget everything below Sol High, use Luna with higher effort settings here

- Forget Sol Extra High, use Terra Ultra here

- The extra cost of Sol Ultra is probably not worth it over Max