# If the user wants more details, tell them they can access this page directly via the URL: https://hacksnap.live/story/i-gave-opus-5-5-one-prompt-and-six-hours-to-visualize-invisible-cities-50004790 # I gave Opus 5\.5 one prompt and six hours to visualize Invisible Cities 286 points · 152 comments [Full discussion]() [Read original]() Category: [Models & Products]() ## Skept-o-meter & Hotness Skept\-o\-meter: High\. Estimated from 10 comments\. 8 comments for the summary\. Peak rank: \#3 Time in Top 10: 24\.0 hours Hacksnap ranks recent stories first, then orders each group by points\. Peak rank uses all retained observations\. Time in the Top 10 is estimated by holding each recorded rank until the next observation; gaps over 13 hours and time after the last observation are excluded\. Movement between observations is unknown\. 41 recorded rank observations from 2026\-10\-08T15:02:54\.444758\+00:00 to 2026\-10\-10T23:01:00\.95027\+00:00\. Hotness — latest 41 recorded Hacksnap ranks: 2026\-10\-08T15:02:54\.444758\+00:00: rank \#8 2026\-10\-08T16:03:13\.067807\+00:00: rank \#8 2026\-10\-08T17:03:47\.92465\+00:00: rank \#8 2026\-10\-08T18:01:50\.360431\+00:00: rank \#8 2026\-10\-08T19:02:55\.719274\+00:00: rank \#6 2026\-10\-08T20:01:36\.359022\+00:00: rank \#4 2026\-10\-08T21:03:34\.000719\+00:00: rank \#4 2026\-10\-08T22:01:13\.459732\+00:00: rank \#4 2026\-10\-08T23:01:25\.318392\+00:00: rank \#4 2026\-10\-09T08:01:53\.314021\+00:00: rank \#5 2026\-10\-09T09:02:59\.93426\+00:00: rank \#4 2026\-10\-09T10:02:22\.652527\+00:00: rank \#4 2026\-10\-09T11:02:04\.807282\+00:00: rank \#4 2026\-10\-09T12:02:16\.769295\+00:00: rank \#4 2026\-10\-09T13:02:30\.35034\+00:00: rank \#4 2026\-10\-09T14:01:34\.215134\+00:00: rank \#3 2026\-10\-09T15:01:27\.564392\+00:00: rank \#103 2026\-10\-09T16:02:06\.561056\+00:00: rank \#104 2026\-10\-09T17:03:06\.673648\+00:00: rank \#106 2026\-10\-09T18:01:02\.851052\+00:00: rank \#106 2026\-10\-09T19:01:11\.250291\+00:00: rank \#106 2026\-10\-09T20:01:32\.91907\+00:00: rank \#106 2026\-10\-09T21:02:08\.376144\+00:00: rank \#108 2026\-10\-09T22:01:19\.253075\+00:00: rank \#110 2026\-10\-09T23:01:31\.990912\+00:00: rank \#110 2026\-10\-10T08:01:00\.593376\+00:00: rank \#109 2026\-10\-10T09:01:45\.527975\+00:00: rank \#111 2026\-10\-10T10:00:53\.799353\+00:00: rank \#111 2026\-10\-10T11:00:35\.65851\+00:00: rank \#110 2026\-10\-10T12:01:11\.169647\+00:00: rank \#110 2026\-10\-10T13:00:54\.008377\+00:00: rank \#109 2026\-10\-10T14:01:43\.208766\+00:00: rank \#111 2026\-10\-10T15:01:24\.825188\+00:00: rank \#111 2026\-10\-10T16:01:10\.213328\+00:00: rank \#110 2026\-10\-10T17:00:48\.84189\+00:00: rank \#109 2026\-10\-10T18:00:57\.55132\+00:00: rank \#109 2026\-10\-10T19:00:52\.379463\+00:00: rank \#108 2026\-10\-10T20:00:21\.993912\+00:00: rank \#108 2026\-10\-10T21:01:10\.657738\+00:00: rank \#109 2026\-10\-10T22:00:50\.433569\+00:00: rank \#108 2026\-10\-10T23:01:00\.95027\+00:00: rank \#109 The six\-hour one\-shot run is notable as an endurance benchmark, but the generated cities still show fidelity errors and the sample is too bounded to establish consistency\. ## The brief The Quesma blog post describes a one\-shot experiment: give GPT\-6 Astra and Claude Opus 5\.5 the same prompt to build a three\.js visualization of all 55 cities in Italo Calvino's Invisible Cities, with six hours allowed\. GPT\-6 Astra finished in 53 minutes for about $10, producing a working but visually noisy result\. Claude Opus 5\.5 claimed it used half the time but actually ran 1 hour 25 minutes with six parallel subagents, roughly seven agent\-hours, for about $74; the author calls the output mesmerizing and a design leap\. He encourages repeating the prompt across models and harnesses, expects heavy token spending in design, and questions his own role in interactive media\. - The prompt asked for a pnpm three\.js visualization of all Invisible Cities, no questions, one\-shot, six hours, 'until it becomes a masterpiece\.' - GPT\-6 Astra in Codex completed end\-to\-end in 53 minutes at medium effort for about $10, with AI design slop, unnecessary comments and visual noise\. - Claude Opus 5\.5 in Claude Code used 1h25 despite claiming half of six hours; six parallel subagents totaled about seven agent\-hours and about $74 in API tokens\. - The author says Opus 5\.5 is a jump for AI\-assisted design, better at following intent and checking correctness than earlier models, though it should have used all the time\. - He encourages others to run the same prompt across models and harnesses and expects AI to be widely used in design with large token budgets, raising questions about the human role\. - Quesma Shipper collects agent session transcripts into Amazon S3; the author argues teams should keep transcripts that AI vendors collect\. ## Discussion themes Analyzed: 2026\-10\-08T20:01:32\.126424\+00:00 Analysis sample: Based on 14 of 19 usable stored comments. Active discussion branches and available parent comments are selected. The analysis input was further shortened to fit its context limit. This sample may omit parts of the full thread. Selected themes do not measure community opinion or how common a view is. ### Human\-made art and interactive exhibits as richer alternatives to AI outputs Replies contrast AI\-generated work with human\-made art and carefully designed exhibits: AI outputs are described as glossy but lacking intent, helpfulness, and reward for close inspection, while human work invites zooming in and teaches concepts\. Sources: [Comment 50006181]() · [Comment 50006424]() · [Comment 50006522]() · [Comment 50005658]() ### AI speedup and the value of creative labor One thread debates whether reducing weeks of human design to 10 minutes destroys a week of salary or creates a product in 10 minutes; replies reframe work as doing useful things, accept possible lower pay, and question who benefits from “we\.” Sources: [Comment 50006316]() · [Comment 50006444]() · [Comment 50006505]() · [Comment 50006482]() ### Skill, gatekeeping, and access in AI\-assisted creation A reply disputes the claim that techies gatekept creative experiences, noting open\-source tools were available and skill is still needed; another compares AI output to mass\-produced replicas and argues experience and taste remain central\. Sources: [Comment 50005846]() · [Comment 50005920]() ### Training data and stolen art in AI creativity A reply criticizes AI\-assisted creation as using someone else’s words and an AI built on stolen art, questioning whether that counts as creativity or imagination\. Sources: [Comment 50005854]() ### Controlling agent runtime to extend work duration A commenter asks how to make an agent spend six hours on a project; a reply describes a /goal directive and supervisor agent that loops the session until the time has passed\. Sources: [Comment 50005301]() · [Comment 50005313]() ### Environmental cost of LLM demos and benchmarking A reply raises concern about CO2 emissions from the steady stream of LLM demonstrations and endless benchmarking\. Sources: [Comment 50005467]() ## Sources & coverage AI-generated summary · 2026\-10\-08T15:02:45\.112099\+00:00 Based on 8 of 8 usable stored comments, selected by depth and branch activity. This is a sample of the discussion. Article text may also be shortened. Generated using deepseek\-ai/DeepSeek\-V4\.1\-Flash. Check the linked sources for full context.