sho

[sho @unlimwallworks]

@jessethanley Game is game

[sho @unlimwallworks]

@sudoingX Hey the worst take I've seen yet

[Theo - t3.gg @theo]

Anthropic has no small models that are worth using right now.

OpenAI has no large models that are worth using right now.

Google has no models that are worth using right now.

[sho @unlimwallworks]

@trq212 Replace it with a super auto mode so I don't have to babysit the agent at all

I accept the consequences

[sho @unlimwallworks]

@VictorTaelin Auto compact at 300k feels great so far

[sho @unlimwallworks]

@opencode Think I'll wait after the union alpha rug pull

[sho @unlimwallworks]

@bridgemindai I think luna and deepseek deserve a tier for just cause of cost, but only applicable for things that aren't coding (Hermes etc)

[sho @unlimwallworks]

@Hesamation Except opus 5.5 is superior to Astra right now, so there's no comparison here.

[sho @unlimwallworks]

@BLUECOW009 You are deeply mistaken.

Opus 5.5 is mind blowingly good. The also increased the 5 hour window. I've been running non stop opus agents and not hit it yet.

If I tried to do the same thing with gpt I would hit my weekly limit instead.

[sho @unlimwallworks]

@theo First you got reserve Luna killed. Now you got Anthopric to release a great model.

[sho @unlimwallworks]

@VadimStrizheus >posts the most factually correct opinion

[sho @unlimwallworks]

@ivanfioravanti 5.5 lives up to the hype, trust me

[sho @unlimwallworks]

@jun_song very bad take on point 2

I switched my Hermes agent from my Mac to Linux cause of persistent and annoying permissions issues

[sho @unlimwallworks]

@theo Luna good
Sol bad

[sho @unlimwallworks]

@MaxRovensky brother we are all having a great time with 5.5, you're the odd one out

[sho @unlimwallworks]

@witcheer Simple mode should be the default honestly. I don't even like how you can drag components as is, should be a setting too.

[sho @unlimwallworks]

@QuestforYield @MiaAI_lab @vSouthvPawv Glm flash felt better in my limited testing

[sho @unlimwallworks]

@synthwavedd I'm tried of benchmarks being (max) reasoning

High is the only benchmark I care about

[sho @unlimwallworks]

@cyrilismyname @bridgemindai lol

[sho @unlimwallworks]

@HarryKarl_ @0xSero I get much more mileage right now out fable than Astra.

Fable is also more consistent, better at orchestration, coding and frontend design