Public video arenas vote only on pre-rendered clips, so no public benchmark yet exists for real-time natural-language steering.
Back to the trend of the day: Google's Genie 4 AI builds worlds to walk through; invited testers only
Poly Rating · from the trend of the day
Genie 4 claims ranked by evidence gap (2026)
Google's own post says Genie 4's worlds can drift after multiple prompt edits, which is its headline feature, and no public benchmark can yet edit a world while someone walks through it.
- No. 1: Genie 4 real-time prompt steering (weather, camera, adding people or objects)
- 10 places
01Final list
The leading AIs each built their own list of ten independently. Here is one result that is easy to check: every place has a reason, a source, and the list of AIs that included it.
Order: from most to least AI agreement.
A bright badge means this AI put the entry on its list. A dim badge means it did not.
- 1
Genie 4 real-time prompt steering (weather, camera, adding people or objects)
This is the headline feature, and Google says it drifts after 'multiple prompt edits'. No open benchmark edits a world mid-walk, and no outside tester has hands-on access.
Included it: Claude, Gemini. Did not include: ChatGPT, Grok, Nova, Kimi, Wenxin, Qwen, StepFun, DeepSeek.In their lists: Claude: 1 · Gemini: see list
2 AIs put it here - 2
Genie 4 long-horizon memory and consistency
Google lists long-term consistency as unsolved. iWorld-Bench's best memory score is 0.85 in the paper summary but 0.91 on the live leaderboard. Genie 4 is untested on either.
Included it: Claude, Gemini. Did not include: ChatGPT, Grok, Nova, Kimi, Wenxin, Qwen, StepFun, DeepSeek.In their lists: Claude: 5 · Gemini: 1
What each AI noted 2
- Claudeplace 5
Two minutes, double Genie 3, but Google's own post says only 'a few minutes' and lists long-term consistency as open research.
- Geminiplace 1
Zero out of 14 tested frontier world models top 0.85 memory symmetry on iWorld-Bench, making Google's seamless return-trip promise the widest unverified gap.
Source ↗
2 AIs put it here - 3
Genie 4 for training AI agents and robots
The agent-training pitch rests on one humanoid-checkers demo (Apptronik Apollo). Neither answer found published sim-to-real transfer numbers.
Included it: Claude, Gemini. Did not include: ChatGPT, Grok, Nova, Kimi, Wenxin, Qwen, StepFun, DeepSeek.In their lists: Claude: 3 · Gemini: 3
What each AI noted 2
- Claudeplace 3
The agent-training story rests on one humanoid-checkers experiment, and one commenter says Genie now looks more like a controllable video generator.
Source ↗ - Geminiplace 3
Positioned as the training engine for autonomous robotics, yet zero published sim-to-real robotic policy transfer data accompanied the showcase.
2 AIs put it here - 4
Genie 4 native audio and interactive dialogue
Google calls it the 'first real-time world model to generate audio natively' and says audio accuracy is still improving. One tester heard a clarinet when the prompt asked for a saxophone.
Included it: Claude, Gemini. Did not include: ChatGPT, Grok, Nova, Kimi, Wenxin, Qwen, StepFun, DeepSeek.In their lists: Claude: 2 · Gemini: 7
What each AI noted 2
- Claudeplace 2
'First real-time world model to generate audio natively', yet one commenter heard a clarinet when the prompt asked for a saxophone.
Source ↗ - Geminiplace 7
Real-time interactive generation of dialogue, footsteps, and spatial acoustics outpaces video-only peers, but latency penalties under streaming loads remain undisclosed.
Source ↗
2 AIs put it here - 5
Genie 4 physics (aerodynamics, collisions, water)
The New Stack's headline says the physics 'still aren't quite right', and Google lists complex physical interactions as unsolved.
Included it: Claude, Gemini. Did not include: ChatGPT, Grok, Nova, Kimi, Wenxin, Qwen, StepFun, DeepSeek.In their lists: Claude: 6 · Gemini: 6
What each AI noted 2
- Claudeplace 6
The New Stack's headline says the physics 'still aren't quite right', and Google lists complex physical interactions as unsolved.
Source ↗ - Geminiplace 6
Navigating environments with keyboard inputs masks persistent object clipping and boundary pass-through errors that formal physical simulators do not allow.
Source ↗
2 AIs put it here - 6
Genie 4 photorealism, lighting and materials
This claim has the smallest gap. Viewers praise the lighting and materials, though one 'photorealistic' clip looked like a video game. Blind arena voting can at least compare offline clips.
Included it: Claude, Gemini. Did not include: ChatGPT, Grok, Nova, Kimi, Wenxin, Qwen, StepFun, DeepSeek.In their lists: Claude: 9 · Gemini: 9
What each AI noted 2
- Claudeplace 9
The most testable claim and the one with the most support: commenters call lighting and materials clearly better than the previous generation.
Source ↗ - Geminiplace 9
The easiest claim to verify visually, backed by high leaderboard placements, yet cinematic lighting contributes least to agent functional competence.
Source ↗
2 AIs put it here - 7
Genie 4 characters and multi-agent behaviour (Gemini Omni grounding)
Google concedes that modelling several independent agents is 'still an open research problem'. The Gemini Omni character grounding appears only in Google's demo reels.
Included it: Claude, Gemini. Did not include: ChatGPT, Grok, Nova, Kimi, Wenxin, Qwen, StepFun, DeepSeek.In their lists: Claude: 7 · Gemini: 5
What each AI noted 2
- Claudeplace 7
Google concedes that modelling interactions among multiple independent agents is 'still an open research problem'.
Source ↗ - Geminiplace 5
Google claims multimodal reasoning models keep non-player character behavior coherent, but multi-agent orchestration routinely de-syncs after rapid branching prompts.
2 AIs put it here - 8
Genie 4 SynthID watermarking
This is a procedural claim, not a performance one, and it will be easy to check once outputs are public. No third party has yet confirmed that the watermark survives in live streams.
Included it: Claude, Gemini. Did not include: ChatGPT, Grok, Nova, Kimi, Wenxin, Qwen, StepFun, DeepSeek.In their lists: Claude: 10 · Gemini: 10
What each AI noted 2
- Claudeplace 10
A procedural claim, not a performance claim, and the easiest to check once outputs are public.
Source ↗ - Geminiplace 10
Google's digital provenance footprint is technically established, facing an evidence gap only regarding its survival after low-latency frame downscaling.
2 AIs put it here - 9
Genie 4 event planning with Gemini 3.8 Flash
A reasoning model plans, checks each step and revises when the world shifts, but no published test measures plan-following inside a generated world.
Included it: Claude. Did not include: ChatGPT, Gemini, Grok, Nova, Kimi, Wenxin, Qwen, StepFun, DeepSeek.In their lists: Claude: 4
What each AI noted 1
- Claudeplace 4
A reasoning model plans, checks each step and revises, but no published test shows it keeps a world coherent.
Source ↗
1 AIs put it here - 10
Genie 4 for film, journalism, education, gaming and engineering
Five tester fields are named, including film 'draft mode' and archival photos turned into worlds, but there are no tester counts or results. Some testers call the output 'fuzzy'.
Included it: Claude. Did not include: ChatGPT, Gemini, Grok, Nova, Kimi, Wenxin, Qwen, StepFun, DeepSeek.In their lists: Claude: 8
What each AI noted 1
- Claudeplace 8
Real use cases are named, but the post gives no tester counts, sample sizes or quantified results.
Source ↗
1 AIs put it here
02Worth knowing
What the AIs noticed while building the list.
Commenters say Genie 4 stays coherent for things in frame and breaks when objects leave and come back. An agent-training simulator needs exactly that case to work.
Gemini handles planning and Genie renders the world, which makes it hard to tell which model deserves credit for each claim.
03How we counted
In plain words, no formulas.
1The question
Top 10 Genie 4 claims ranked by evidence gap
2Where it applies
global AI world models
3What we looked at
- Strength of the public claim in Google’s announcement and demos
- Independent coverage: what reporters say testers saw versus what Google states
- Benchmark coverage: whether existing open tests can even measure the claim
- Failure consequence: how damaging the admitted limitation is if it shows up in real workflows
- Google DeepMind: Genie 4 official post — The claims and Google's own caveats (drift, object permanence, audio accuracy, multi-agent interaction). It gives no resolution, frame rate, session length or
- Quartz launch report — Who has access (trusted testers only), keyboard movement, how Gemini Omni and Gemini reasoning models fit in, and the Bloomberg-reported limit of about two minutes per simulatio
- SiliconANGLE launch report — Rollout plan and the two-minute consistency window, described as double Genie 3's limit
- Android Authority launch report — What reporters could actually see: Google's own videos, restricted access, no stated resolution
4Who ranked
The leading AIs, each building its own list of ten independently and naming a source for every place: Claude, ChatGPT, Gemini, Grok, Nova, Kimi, Wenxin, Qwen, StepFun, DeepSeek.
5How the places were set
The same entry written differently by different AIs was counted as one. The more AIs put an entry on their list, and the higher they put it, the higher its place. "AI agreement" shows how many of the AIs had it on their own list. Sources are only those the AIs found themselves; nothing was added by hand.
04Each AI's own list
🇺🇸 Claude10 places · 3 notes
The same question, answered by Claude alone. The final list above combines all the AIs.
- 1Real-time natural-language steering of a live world (weather, camera, adding people or objects)
The headline feature is the one Google itself says breaks after 'multiple prompt edits', and no public test edits a world mid-walk.
Source ↗ - 2Interactive dialogue that adapts to user input (native audio)
'First real-time world model to generate audio natively', yet one commenter heard a clarinet when the prompt asked for a saxophone.
Source ↗ - 3Usefulness for training agents and robots (SIMA, Gemini Robotics, Apptronik Apollo checkers)
The agent-training story rests on one humanoid-checkers experiment, and one commenter says Genie now looks more like a controllable video generator.
Source ↗ - 4Complex event simulation guided by Gemini 3.8 Flash
A reasoning model plans, checks each step and revises, but no published test shows it keeps a world coherent.
Source ↗ - 5Long-horizon consistency and memory (about two minutes)
Two minutes, double Genie 3, but Google's own post says only 'a few minutes' and lists long-term consistency as open research.
- 6Physics simulation (aerodynamics, collisions, flowing water)
The New Stack's headline says the physics 'still aren't quite right', and Google lists complex physical interactions as unsolved.
Source ↗ - 7Characters that move, react and behave naturally (including multi-agent interaction)
Google concedes that modelling interactions among multiple independent agents is 'still an open research problem'.
Source ↗ - 8Trusted-tester value in film, journalism, education, gaming and engineering
Real use cases are named, but the post gives no tester counts, sample sizes or quantified results.
Source ↗ - 9'Most realistic' photorealism, lighting and materials
The most testable claim and the one with the most support: commenters call lighting and materials clearly better than the previous generation.
Source ↗ - 10SynthID watermarking on all generated video and audio
A procedural claim, not a performance claim, and the easiest to check once outputs are public.
Source ↗
What Claude noticed
Google's own post lists no resolution, frame rate, session length or tester count. The most concrete number in the coverage, two minutes, comes from press reports.
Hacker News commenters say Genie 4 holds coherence well for things that stay in frame and breaks when objects leave and re-enter. That is exactly the case an agent-training simulator needs.
Part of the pitch is a hybrid design: Gemini handles structure and planning while Genie renders the look, sound and response. That makes the claims about Gemini models and Genie hard to separate.
🇺🇸 Gemini10 places · 2 notes
The same question, answered by Gemini alone. The final list above combines all the AIs.
- 1Long-Horizon Memory and Loop-Closure Symmetry
Zero out of 14 tested frontier world models top 0.85 memory symmetry on iWorld-Bench, making Google's seamless return-trip promise the widest unverified gap.
Source ↗ - 2Live Natural-Language Steering and Scene Mutation
Pairwise arenas like Artificial Analysis evaluate static prompt execution, leaving mid-flight interactive prompt steering completely unbenchmarked by third parties.
Source ↗ - 3Scalable Embodied Agent Training and Sim-to-Real Transfer
Positioned as the training engine for autonomous robotics, yet zero published sim-to-real robotic policy transfer data accompanied the showcase.
- 4Two-Minute Continuous Simulation Stability
While Google highlights open-ended sandbox play, independent reporter briefings revealed a hard boundary of roughly 120 seconds before structural hallucination sets in.
- 5Dynamic NPC Dialogue Grounding via Gemini Omni
Google claims multimodal reasoning models keep non-player character behavior coherent, but multi-agent orchestration routinely de-syncs after rapid branching prompts.
- 6First-Person 3D Geometry and Collision Physics
Navigating environments with keyboard inputs masks persistent object clipping and boundary pass-through errors that formal physical simulators do not allow.
Source ↗ - 7Native Synchronized Multi-Channel Spatial Audio
Real-time interactive generation of dialogue, footsteps, and spatial acoustics outpaces video-only peers, but latency penalties under streaming loads remain undisclosed.
Source ↗ - 8Responsive Keyboard Latency and Camera Tracking
Testers confirmed functional WASD navigation, but end-to-end cloud frame-rendering latency fluctuates significantly outside local high-bandwidth rigs.
- 9Photorealistic Lighting, Weather, and Volumetric Materials
The easiest claim to verify visually, backed by high leaderboard placements, yet cinematic lighting contributes least to agent functional competence.
Source ↗ - 10SynthID Watermarking Across Interactive Audio and Video
Google's digital provenance footprint is technically established, facing an evidence gap only regarding its survival after low-latency frame downscaling.
What Gemini noticed
The memory ceiling is the defining technical barrier of 2026: iWorld-Bench tests show no interactive world model exceeds 0.85 memory symmetry, explaining why DeepMind's demo clips avoid extended loop-closure navigation.
Public leaderboards like the Artificial Analysis Video Arena are structured for non-interactive video pairwise voting, leaving the entire industry without a public benchmark for real-time natural language steering.