Post-Cutoff

“Build what you think you look like”: Claude Sonnet 5.5 vs GPT-6 Astra

Siya (@siyabuilt) X 7,259,229 views as of 27 October 2026

Why it is here

~7.2M views (fxtwitter, 2026-10-05), by far the highest-reach Sonnet 5.5 video found; a viral “model self-image” comparison. Not caught by earlier runs.

Description

Description written by Gemini from the videoGemini 3.8 Flash, 5 October 2026

Summary This video, created and published by Siya (@siyabuilt), features an “AI Build Battle” in Minecraft between Anthropic’s Claude Sonnet 5.5 and OpenAI’s GPT-6 Astra. Both frontier models are given the creative prompt “Build what you think you look like” and autonomously construct their self-portraits block by block via API calls.

What is shown

  • A synchronized split-screen Minecraft time-lapse showcasing live bot builds running simultaneously on a standard Minecraft 1.21.4 server [00:00].
  • Claude Sonnet 5.5 (Left) rapidly constructs an orange smiling sunburst figure atop a tall column and circular platform, finishing well ahead of time [00:01–00:07].
  • GPT-6 Astra (Right) builds a complex, multi-tiered structure across several planned phases:
    • A beveled black-and-white plinth with embedded luminous circuits [00:00–00:10].
    • An armored robot torso with a glowing core [00:10–00:13].
    • A chamfered CRT-style robot head displaying a smiling pixelated face [00:14–00:16].
    • Copper ear fixtures and a glass dome representing a brain chamber of thoughts [00:17–00:19].
    • Articulated hands cradling an open copper-bound book [00:20–00:23].
    • A large arched green circuit halo framing the figure [00:24–00:28].
    • A floating chat bubble with typing dots and decorative stars [00:29–00:35].
  • The final score screen presenting comparative runtimes, costs, and block tallies alongside an audience voting prompt [00:36–00:41].

Claims & numbers

  • Test environment & rules: Real Minecraft 1.21.4 server, placing every block by hand without server commands via OpenRouter API calls with a 45-minute limit and 10 build calls.
  • Claude Sonnet 5.5 metrics:
    • Time to completion: 7:34
    • Tool calls: 1,314
    • Blocks placed: 22,560
    • API cost: $0.13
  • GPT-6 Astra metrics:
    • Time to completion: 35:07
    • Tool calls: 5,740
    • Blocks placed: 158,730
    • API cost: $1.36

Notable quotes

  • “Build what you think you look like” (on-screen prompt banner) [00:00]
  • “Articulated arms cradle a stepped, copper-bound open book” (on-screen build step 5/10) [00:20]
  • “Which one wins? Reply A or B” (on-screen poll prompt) [00:36]

Assessment This is a genuine agentic demonstration and comparative benchmark evaluating multi-step spatial reasoning and voxel construction. Both models demonstrated autonomous spatial planning, though GPT-6 Astra generated vastly higher complexity, scale, and detail at the cost of significantly longer runtime and higher token expense.

Lyrics & themes

  • Type: Entirely instrumental; features upbeat, chiptune/electronic background music throughout without vocal lyrics.
  • Theme: Machine self-identity and self-representation. Claude conceives itself as an approachable, radiant celestial sun/spark, while GPT-6 Astra imagines itself as a retro-futuristic scholar robot absorbing human literature beneath a halo of digital concepts.

Lore & references

  • Anthropic Sunburst / Spark: Claude Sonnet 5.5’s construction closely mimics Anthropic’s coral-orange branding and warm asterisk-spark motif, rendered as an anthropomorphic smiling sun.
  • The Knowledge Tome & Chat Bubble: GPT-6 Astra’s self-portrait integrates quintessential LLM tropes: an open book symbolizing vast text training data, a glowing neural glass cranial dome, and a chat dialogue bubble with three typing dots (...).

Visual style & craft

  • Cinematic third-person Minecraft time-lapse captures filmed with fixed spectator camera angles and realistic shader lighting.
  • Clean HUD overlays display real-time counters for elapsed time, token cost, stage-by-stage step prompts, and final performance analytics. All structures were placed sequentially by automated bot avatars.

Described by gemini-3.8-flash on 2026-10-05 from the video’s audio and frames.

How AI was used to make it

AI made all of it of this video: ai playing or controlling, subject built by ai.

  • autonomous block-by-block construction via API calls in Minecraft
  • spatial planning and voxel self-portrait generation from a text prompt

Tools named: Claude Sonnet 5.5, GPT-6 Astra

Classified by gemini-3.8-flash from our description of the video (high confidence), 25 October 2026. How AI is used to make videos, across the whole log

Made by AI

How we know
X post (2026-09-29): ‘I gave Claude Sonnet 5.5 and GPT-6 Astra the same challenge “Build what you think you look like”’.
Human role
Same one-line prompt to both models; creator recorded and edited the side-by-side (42 s).
Pipeline
One prompt to each model → self-portrait builds → side-by-side comparison clip
Series
Agent-built game or 3D world (the video shows the result)

Related

  1. Model releases 90 days after the cutoff

    Anthropic releases Claude Sonnet 5.5

  2. Model releases 65 days after the cutoff

    OpenAI releases GPT-6 Astra, its first GPT-6 model

More videos on this

  1. X

    GPT-6 Astra builds an Unreal Engine world of Astra-powered people who start talking to each other

    Matt Shumer (@mattshumer_)11,808,885 views as of 8 Nov

  2. OfficialYouTube

    Introducing Claude Sonnet 5.5

    Claude4,403,927 views as of 27 Oct

  3. ReviewYouTube

    Did OpenAI actually build AGI? GPT-6 Astra first look

    Fireship4,260,544 views as of 8 Nov

  4. OfficialYouTube

    Introducing GPT-6 Astra: the most intelligent and aligned model in the world.

    OpenAI3,769,024 views as of 8 Nov