Vas

Premium launch videos for ambitious startups and established brands - built to make products feel memorable and impossible to ignore.

[Vas @CultofVas]

@joshwoodward I remember they were ahead in most benchmarks during other releases too like 3.1 pro

[Vas @CultofVas]

@arena @GoogleDeepMind I remember they were ahead in benchmarks during other releases too

[Vas @CultofVas]

@GoogleDeepMind This has to be a meme

[Vas @CultofVas]

@MysteryEthos @thsottiaux Try buying $100 of API credits after 5x is over you will feel the incentives

[Vas @CultofVas]

Crazy how Sol went from a Fable competitor to a Sonnet equivalent in just a few weeks

[Claude @claudeai]

Introducing Claude Sonnet 5.5, the second model in the Claude 5.5 family.

It’s a clear upgrade over Sonnet 5, runs more than 30% faster, and costs up to 30% less for most work.

[Vas @CultofVas]

@madebypuru @goatfishxyz Bait posts like this make potential clients willing to pay $$$$$ think they can get the same result in Opus for just $$.

[Vas @CultofVas]

They aren’t real, but even the actual creators were ragebaiting before, posting genuinely high-quality work with captions like “Created this with 5 prompts” just to drive engagement. That made potential clients believe this level of work could be achieved easily with AI, and now, after opus 5.5, many of them genuinely think it can, even though that's wrong.

[Vas @CultofVas]

@anshnanda API cost would be high in that case, right?

[Vas @CultofVas]

@runwayml This is why they are building extensions to filter slop out from feeds

[Vas @CultofVas]

@alisaqqt Was wondering what happened to Kling. 3.0 is still good though.

[Vas @CultofVas]

@lmstudio @inco_ai Working well on my M5 Max 64 GB. I haven't tried anything intensive, though.

[Vas @CultofVas]

Testing GPT Images 2.5 in real video projects already.

[Vas @CultofVas]

@higgsfield GPT models only give the still frame, how does the animation comparison work?

[Vas @CultofVas]

@FNFFAN737 Because 90% of the engagement comes from that ragebait

[Vas @CultofVas]

GTA 6 mods are gonna be crazy

[Matt Shumer @mattshumer_]

GPT-6 Astra built this Manhattan world in Unreal Engine over the course of a week.

It was literally able to go street by street to make each one perfect.

[Jun Song @jun_song]

Current state of AI models:

Anthropic: Most advanced performance, ridiculously expensive
OpenAI: Mid models, overhyped and overpriced
xAI: Weak models, overhyped but slightly cheaper
Google: Trains their models on magic mushrooms

[Vas @CultofVas]

@godswayfoundinc I heard it is not great for Macs because they lack tensor cores. Which Topaz model do you use the most? Worth it for an M5 Max 64 GB?

[Vas @CultofVas]

@avijeet_writes Vantara Creamery 🫡

[Vas @CultofVas]

@diffractstudio We get to see new cool slop creatures

[Vas @CultofVas]

@jun_song 4-bit or 8-bit Qwen 3.8 for M5 Max 64 GB?