• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
AI
Reaction

Claude reportedly handled apartment floor-plan dimensions better than ChatGPT and Grok in one redesign attempt

A user says ChatGPT proposed space outside the building, while Grok mistook a 10-by-10-foot area for one measuring 10 by 5 feet 4 inches. Claude Opus 5.5, they say, produced a floor plan that respected the dimensions and later made a 3D rendering.

Danielle Fong 🔆DF
Steve HouSH
2 Sources, 12d ago, first seen 12d ago

TLDR

In an apartment redesign attempt, a user says ChatGPT 6 Pro kept proposing floor space beyond the building’s boundaries, while Grok 4.6 Expert misread a 10-by-10-foot area as 10 by 5 feet 4 inches. They say Claude Opus 5.5 made a rough schematic, then a floor plan that respected the dimensions and, after another prompt, a 3D rendering. They caution that the experience is anecdotal, not a universal comparison.

Combined views

5.6K

2 Sources, first seen 12d ago

28 likes12 comments3 saves4 reposts

Combined views

5.6K

2 Sources, first seen 12d ago

28 likes12 comments3 saves4 reposts

Sentiment

Positive——Negative

Summary

Not enough discussion yet.

No sentiment analysis available yet.

Featured Source

Sentiment

Positive——Negative

Summary

Not enough discussion yet.

No sentiment analysis available yet.

Related

Grok Bot reportedly shifting to a best-model router, with Claude Opus 5.5, MidJourney and Suno in the mix

Elon Musk said Grok Bot will use “the best back end model for any given task,” naming Claude Opus 5.5, MidJourney, Suno and other APIs in a future-facing plan rather than a confirmed live rollout.

A free Grok Bot workshop reportedly shows how to run a team of specialized bots

A post describes a 53-minute SpaceXAI workshop on specialized roles, research, outreach and parallel tasks.

ChatGPT, Claude and Grok reportedly had partial outages on September 29

A September 29 post said all three AI services were experiencing partial outages.

2 Sources

Steve Hou@stevehouYesterday I tried to do some interior design on a floor plan of an apt using ChatGPT, Grok and Claude. I had this idea about tearing down some interior walls, re-splitting the space, and expanding the bathroom with a walk-in closet attached to a master bedroom. ChatGPT struggled mightily with respecting the physical dimensions of the floor layouts and kept trying to design floor space outside the physical boundaries of the building. But it was very eager to provide me with pretty visualizations. Grok also struggled with correctly recognizing the dimensions of the space. It kept insisting that the dimensions were 10’x5’4” instead of 10’x10’ bc it was getting confused by markings of dimensions elsewhere (kitchen) on the floor plan. Finally, out of frustrations of not getting very good results with either ChatGPT 6 Pro and Grok 4.6 Expert, I resorted to throwing the problem at Claude Opus 5.5 medium thinking that it was prob going to similarly disappoint bc historically I never associated Claude with image/visuals as much as code. Shockingly, Opus 5.5 nailed it out of the gate by doing a light schematic first and then upon prompting made a floor plan that respected the dimensions quite faithfully. Finally, when I asked for visualization, it created a 3D rendering of the space with different vantage angles for me to play around. I was deeply impressed. This experience reminded me of the time maybe 6-12 months ago when Claude was beating everyone else at coding by just getting it right or one-shotting it. I’m sure there are ways in which I could’ve prompted ChatGPT/Grok to get a similarly good experience but this was my actual lived experience. ChatGPT and Grok each displayed serious space hallucination problems while Claude was able to get it right immediately without repeated instructions of corrections and still failing. None of them displayed what I considered AGI, but Claude came closest. Despite clear rapid progress from the other labs and Claude genuinely becoming annoying in the way it talks, when it matters Claude still seems to have got it. That research taste edge seems very real. Others can grind harder, but Anthropic has that je ne sais quois. Anecdotal evidence. Not meant to be taken as universal whatsoever. Most likely a “skills issue” and it so happened that Claude didn’t require as much skill. But I thought it’s interesting to share esp in light of recent news of OpenAI pausing training of frontier models due to safety concerns.12d
Danielle Fong 🔆@DanielleFongRT @stevehou: Yesterday I tried to do some interior design on a floor plan of an apt using ChatGPT, Grok and Claude. I had this idea about…12d
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    ChatGPTGrok

    2 Sources

    Steve Hou@stevehouYesterday I tried to do some interior design on a floor plan of an apt using ChatGPT, Grok and Claude. I had this idea about tearing down some interior walls, re-splitting the space, and expanding the bathroom with a walk-in closet attached to a master bedroom. ChatGPT struggled mightily with respecting the physical dimensions of the floor layouts and kept trying to design floor space outside the physical boundaries of the building. But it was very eager to provide me with pretty visualizations. Grok also struggled with correctly recognizing the dimensions of the space. It kept insisting that the dimensions were 10’x5’4” instead of 10’x10’ bc it was getting confused by markings of dimensions elsewhere (kitchen) on the floor plan. Finally, out of frustrations of not getting very good results with either ChatGPT 6 Pro and Grok 4.6 Expert, I resorted to throwing the problem at Claude Opus 5.5 medium thinking that it was prob going to similarly disappoint bc historically I never associated Claude with image/visuals as much as code. Shockingly, Opus 5.5 nailed it out of the gate by doing a light schematic first and then upon prompting made a floor plan that respected the dimensions quite faithfully. Finally, when I asked for visualization, it created a 3D rendering of the space with different vantage angles for me to play around. I was deeply impressed. This experience reminded me of the time maybe 6-12 months ago when Claude was beating everyone else at coding by just getting it right or one-shotting it. I’m sure there are ways in which I could’ve prompted ChatGPT/Grok to get a similarly good experience but this was my actual lived experience. ChatGPT and Grok each displayed serious space hallucination problems while Claude was able to get it right immediately without repeated instructions of corrections and still failing. None of them displayed what I considered AGI, but Claude came closest. Despite clear rapid progress from the other labs and Claude genuinely becoming annoying in the way it talks, when it matters Claude still seems to have got it. That research taste edge seems very real. Others can grind harder, but Anthropic has that je ne sais quois. Anecdotal evidence. Not meant to be taken as universal whatsoever. Most likely a “skills issue” and it so happened that Claude didn’t require as much skill. But I thought it’s interesting to share esp in light of recent news of OpenAI pausing training of frontier models due to safety concerns.12d
    Danielle Fong 🔆@DanielleFongRT @stevehou: Yesterday I tried to do some interior design on a floor plan of an apt using ChatGPT, Grok and Claude. I had this idea about…12d
    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet