Positive users praise the outlined multi-model AI workflow for managing big projects because models like SOL and FABLE deliberate well and force clarity, while negative users call it convoluted with context lost across handoffs.
Based on 10 visible X reactions from 29 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@swyx @mattpocockuk @trq212 Every model handoff loses context. Four models to plan, critique, code, review means you spend your time reconciling four opinions instead of shipping. For most work one strong model end to end beats the relay race. The stack becomes the bottleneck.
@swyx @mattpocockuk @trq212 I’m seeing tons of posts to avoid sol ultra for draining tokens but I haven’t had any issues with it. Totally agree with using fable to review sol’s choices
@swyx @mattpocockuk @trq212 dumb q but this feels kinda convoluted honestly? what are the advantages of using this over just sol ultra for everything? saving tokens?
@swyx @mattpocockuk @trq212 Have you tried Grok 4.5? It is working pretty well for me so far!
@swyx @mattpocockuk @trq212 Every model handoff loses context. Four models to plan, critique, code, review means you spend your time reconciling four opinions instead of shipping. For most work one strong model end to end beats the relay race. The stack becomes the bottleneck.
@swyx @mattpocockuk @trq212 I’m seeing tons of posts to avoid sol ultra for draining tokens but I haven’t had any issues with it. Totally agree with using fable to review sol’s choices
@swyx @mattpocockuk @trq212 dumb q but this feels kinda convoluted honestly? what are the advantages of using this over just sol ultra for everything? saving tokens?
@swyx @mattpocockuk @trq212 Swap sonnet 5 for Grok 4.5. Trust me
@swyx @mattpocockuk @trq212 Fantastic
where i'm currently at for Big Boy projects: - sol ultra to plan - fable 5 to critique - sonnet 5/terra ultra/swe 1.7 to ultracode/slop cannon - devin review to review (using kakuna) ~always use a variant of @mattpocockuk's grill-me or @trq212's interview-me to elicit decisions upfront
@mattpocockuk @trq212 what other people are recommending to use https://www.latent.space/p/ainews-not-much-happened-today-f5c
Positive users praise the outlined multi-model AI workflow for managing big projects because models like SOL and FABLE deliberate well and force clarity, while negative users call it convoluted with context lost across handoffs.
Based on 10 visible X reactions from 29 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@swyx @mattpocockuk @trq212 Fantastic
where i'm currently at for Big Boy projects: - sol ultra to plan - fable 5 to critique - sonnet 5/terra ultra/swe 1.7 to ultracode/slop cannon - devin review to review (using kakuna) ~always use a variant of @mattpocockuk's grill-me or @trq212's interview-me to elicit decisions upfront
@mattpocockuk @trq212 what other people are recommending to use https://www.latent.space/p/ainews-not-much-happened-today-f5c