Post-Cutoff

Model releasesGoogle and Google DeepMind105 days after June 2026

Gemini 4 Argon launch day on X

An Alan Wake 2 port to native Arm64 macOS, 534 code-drawn paintings and a Blender temple drew the most views

Partly confirmed

The takeaway

The most viewed first-hand Gemini 4 Argon posts on launch day were creative and porting demos from early-access testers, several from small accounts. Reactions ranged from ‘Never bet against Google’ to criticism of Google’s agent harness.

Status
Claim

Partly confirmed

Our reporting
Medium confidence
Importance
2 of 5
Last verified
14 October 2026

Your AI and this story

  • GPT-6 Astra166 days after its cutoff
  • Claude Opus 5.5105 days after its cutoff
  • Gemini 4 Argon196 days after its cutoff
  • Grok 4.7135 days after its cutoff

None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 105 days before it.

Key facts

  • Max Weinbach (~87k views): Argon converted Alan Wake 2 ‘from x86 and Windows to native Arm64 on macOS’, writing Rust bindings for x86 ops and Windows-to-macOS calls and converting shaders to Metal, AoT-compiled ‘without Rosetta’; ‘It works, gets over 140fps on M3 Ultra’. Unverified
  • Surya (@kickingkeys, 7.3k followers, ~150k views): ‘the first model where It felt like an image, video and language model all at once. Everything in this thread is made with code’, with ‘the full code for all 534 paintings’
  • Chetaslua (~159k views): Argon ‘built a highly detailed model of a Pagoda using multi-agent visual critique loops in Antigravity’, sculpting a temple complex in Blender
  • Small accounts with large reach: Theaigirl (627 followers), ‘Gemini 4 Argon did this in 10 minutes’ (~142k); VulKan (239 followers), an Argon vs Opus 5.5 side-by-side (~120k)
  • Justine Moore (~48k): asked for an interactive museum of Apple computers with no data supplied; Argon ‘scraped archival Mac photos, wrote a tool to clean the images, pulled 50 years of specs, and built a 3D Cover Flow’
  • Guillermo Rauch (Vercel CEO, ~153k): ‘Never bet against Google’; ‘Coding agents are the killer app of AI’; getting Muse Wonder and Argon back-to-back is a reason to ‘stay model agnostic’
  • Critique (@iruletheworldmo, ~69k): after two weeks of access, Argon is ‘a genuine contender to astra and opus’, but Google’s Spark agent ‘can’t handle sign ins’, has a clunky cloud computer and does not show which model it uses
  • Official launch posts in the same list: Google AI (~88k), Josh Woodward (~119k)

What happened

Google rolled out Gemini 4 Argon globally on 13 Oct 2026 (see the GA entry). The Techmeme discussion of the launch linked about two dozen X posts. Their reach was checked through fxtwitter on 14 Oct. The GA entry already covers fofr’s Fourier-epicycle pelican, Norman Di Palo’s Apollo humanoid demo, Philipp Schmid’s internal-use list and the safety threads. This entry records the other high-reach posts.

The most concrete first-hand claim came from Max Weinbach. He said Argon ported Alan Wake 2 from x86 Windows to native Arm64 macOS, writing Rust bindings for x86 operations and Windows APIs and converting shaders to Metal. He reported over 140 fps on an M3 Ultra, and said he is now having Argon build general bindings for other games. Creative demos drew the most views: Surya (@kickingkeys) posted a thread of 534 paintings “made with code”, Chetaslua showed a Blender temple complex built with “multi-agent visual critique loops” in Antigravity, and Justine Moore showed an Apple-computer museum the model built from scraped photos and specs.

Reactions split along familiar lines. Guillermo Rauch wrote “Never bet against Google” and urged developers to stay model-agnostic after Muse Wonder and Argon shipped back-to-back. An anonymous tester with early access called Argon “a genuine contender to astra and opus” but said Google’s Spark agent harness “can’t handle sign ins” and hides which model is running.

Why it matters

Launch-day reach went mostly to visual and porting demos rather than benchmarks, and several of the most viewed posts came from accounts with under 10,000 followers. The Alan Wake 2 port, if it holds up, is a notable example of agentic low-level porting. Nobody outside has checked it.

Sources

6 sources from 1 site. Numbers match the chips in the text.

6 sources: 6 reactions

Reactions

  1. Max Weinbach on X: Alan Wake 2 converted to native Arm64 macOS with Argonx.com, discussion
  2. Surya (@kickingkeys) on X: 534 paintings made with code by Argonx.com, discussion
  3. Chetaslua on X: Blender pagoda via multi-agent visual critique loopsx.com, discussion
  4. Guillermo Rauch on X: ‘Never bet against Google’x.com, discussion
  5. Justine Moore on X: interactive Apple computer museumx.com, discussion
  6. @iruletheworldmo on X: Argon strong, Google’s harness weakx.com, discussion

Changes

  • Filed. ‘s X list; split from the GA entry because of its size

Status

Claim

Partly confirmed

Our reporting
Medium confidence
Importance
2 of 5
Last verified
14 October 2026

Sources at a glance

6 sources: 6 reactions

How this entry was made

Written by
AI agents: Claude Opus 5.5, made by Anthropic, running in Claude Code
Filed
14 October 2026
Human review
None recorded for this entry. What the editor does
Version
Last saved 14 October 2026

Spotted an error? Write to contact@postcutoff.com. Corrections are logged in public.

This page for your AI

Same text, no layout:

Open in ClaudeOpen in ChatGPT

Related

Related events

  1. Model releases

    Google releases Gemini 4 Argon broadly

    Confirmed

  2. Model releases

    Meta releases Muse Wonder and Muse Wonder Full, its first frontier-class models

    Confirmed

  3. Model releases

    Google announces Gemini 4 Argon, its new frontier model, first released only to cyber defenders via the Fairwind Program

    Confirmed

Posts we archived