Soul Framer

Creating advance AI automation tools.
Currently working on a Harness.

[Soul Framer @soulframer]

@SpaceXAI 21 voices in one update is a big jump, can't wait to try the range.

[Soul Framer @soulframer]

@TencentHunyuan Hy3 clearly outperforms the rest in all 3 scenes, check the atomic chat comparison below.

https://t.co/KptW1pDgct

[atomic.chat @atomic_chat_hq]

New Hunyuan Hy3 hits Gemini 3.5 quality on physics for 35x cheaper!

We gave 4 models the same prompt: build three self-contained HTML5 canvas scenes with real physics demos

Prompts:

- A bowling ball knocking down the pins
- An air hockey rally that ends in a goal
- A pool break scattering the rack

Outputs:
Hunyuan Hy3: 29,757 tokens, $0.006
Gemini 3.5: 23,300 tokens, $0.21
GLM-5.2: 25,454 tokens, $0.07
DeepSeek-V4: 50,600 tokens, $0.009

Tencent's Hy3 matched Gemini across all three: clean collisions, the puck bounced true, the pins scattered like a real strike, the rack broke with real momentum, nothing clipped or floated. GLM is genuinely strong on pure coding tasks, but the moment the job steps outside clean code it gives way. DeepSeek was the letdown, it burned the most tokens of anyone (50k, almost 2x Hy3) and still turned in the weakest scenes

[Soul Framer @soulframer]

@atomic_chat_hq collisions and momentum in all 3 scenes is impressive, Hy3 surely outperforms.

[Soul Framer @soulframer]

@xai great to have everything in one interface at $0.05/min, more production ready.

where as a couple of days ago hermes + elevenlabs collab needs more setup and managing separate resources.

[Soul Framer @soulframer]

@NousResearch great updates.
personally I like /learn - turn anything into a reusable skill is useful.

[Soul Framer @soulframer]

@tonysimons_ @Teknium This is fun 😂

[Soul Framer @soulframer]

@aiedge_ connecting business ops to automate everything is really cool.

[Soul Framer @soulframer]

@aiedge_ for someone starting form scratch, how much time will it take to build the workflow?

[Soul Framer @soulframer]

Hermes now let's you merge any two AI's, like GPT 5.5 + Claude Opus into one virtual Mixture of Agents model.

The results, 8% higher than Opus 4.8 and 11% higher than GPT 5.5. Full features intact.

[Nous Research @NousResearch]

The strongest models are gated and access is granted only to a select few.

Hermes Agent now exposes MoA presets as virtual models, giving you capabilities beyond the publicly available frontier: 8% higher than Opus 4.8 and 11% higher than GPT 5.5 on our upcoming benchmark.

[Soul Framer @soulframer]

@NousResearch nice one...turning MoA presets into virtual models and making them available for regular users. Hermes team :)

[Soul Framer @soulframer]

@atomic_chat_hq the real challenge would be turning a prescription, student's messy notes or quick session MOM into structured text and if it's long summarize it.

[Soul Framer @soulframer]

@ElevenLabsDevs agent finally has a voice.

can't write?
just talk.

[Soul Framer @soulframer]

@NousResearch the way it rejects the first few ideas and looks for a method for the constraint is really creative, by not capturing the tree and capture what it leaves behind.

Shadow. Texture. The tree is never in the frame and somehow the story exists. I like the BGM too.

[Soul Framer @soulframer]

@NousResearch skills without memories is still stateless.

the "learn from anything" direction is amazing.

a character that remembers "who it is" across every conversation, that's what I am working on.