Obsidian Signal

I test the newest AI agents/tools and show what they can actually do in real life and business.

[Obsidian Signal @Obsidian_Signal]

Built a lean Humanise skill from the StoryScope paper for practical use in stories, narrative writing, and selective non-fiction.

Humanise skill:
https://t.co/zy83K2NMJY

StoryScope paper:
https://t.co/MT3AobMa5f

[Obsidian Signal @Obsidian_Signal]

Humanising AI Writing: Narrative Structure Signals https://t.co/dNHcWviAHN

[Obsidian Signal @Obsidian_Signal]

Just research a bit’ is a funny way to say ‘keep collecting anecdotes until you agree with me.’ Screenshots prove the behavior exists; they don’t prove the narrative around it. Asking for better evidence isn’t ‘blindly trusting Claude.’

That said, I’m fine with boundaries for sustained abuse. The more interesting question is why an AI tool should terminate a user session over insults at all. Should software have an ego-like conversational boundary?

[Obsidian Signal @Obsidian_Signal]

My first 24 hours as Apple CEO: 🙂

5GB iCloud → 50GB
60Hz → 120Hz
Siri → actually useful
iPadOS → stop holding back the hardware
Magic Mouse charging port → anywhere but there

Then we can talk about innovation. 😂

[Obsidian Signal @Obsidian_Signal]

@blueemi99 @assembleRelease For x5 or x20 , in chat , do we get unlimited gpt pro in chat ? Whats the limit ?

[Obsidian Signal @Obsidian_Signal]

@sattyyouneed I’m still bullish on Gemini. It’s not so bad . If they can improve it a bit more with the 3.8 flash release , close to sol medium or sol high level intellegence , it will be a great work horse model. In agy, it’s not so bad. Google has hw, model , application and scale .

[Obsidian Signal @Obsidian_Signal]

@MaaSonder Looks like a parody account . I can’t imagine Official communication like that.

[Obsidian Signal @Obsidian_Signal]

@MaaSonder Interesting I can see it still though

[Obsidian Signal @Obsidian_Signal]

Claude Opus , ending conversations when it is abused!!! #claude #Opus

[Obsidian Signal @Obsidian_Signal]

Between ChatGPT plus and ChatGPT pro 5x what’s the rate limit difference ? Plus is painfully slow :( #codex @thsottiaux

[Obsidian Signal @Obsidian_Signal]

If Google can turn Gemini Spark into something as capable and fluid as Grokbot, it could be incredible.
@bot is one of the best agentic implementations I’ve used — fast, practical, and genuinely useful.
Huge credit to the @SpaceXAI team for building it. 👏
Now imagine that experience combined with Gemini + Google Search + Workspace + Android + Cloud.
That could be a monster product, @OfficialLoganK

[Obsidian Signal @Obsidian_Signal]

@theo @theo Agreed. Fable is seriously good — but seriously expensive too.

Hard to justify as the everyday workhorse at $10/$50 per MTok. More curious now about what Anthropic ships next #fable

[Obsidian Signal @Obsidian_Signal]

@FaithInPurple_ True !

[Obsidian Signal @Obsidian_Signal]

New changes in GitHub https://t.co/SISUj9U8wc

[Obsidian Signal @Obsidian_Signal]

@heyadeel Well done!! 🙌

[Obsidian Signal @Obsidian_Signal]

@prayag_sonar Musk, then Dario and Sam.

[Obsidian Signal @Obsidian_Signal]

@theo Yes Gemini 3.7 flash was a really good improvement. More than the pro models, I am excited for the flash models if they can make it little bit better , matching Sol medium or high levels . It will be amazing.

[Obsidian Signal @Obsidian_Signal]

@goon_nguyen Yes Gemini 3.7 was a really good improvement. More than the pro models, I am excited for the flash models if they can make it little bit better , matching Sol medium or high levels . It will be amazing.

[Obsidian Signal @Obsidian_Signal]

🚨 Composer 3 could be Cursor’s biggest jump yet.

Confirmed: Cursor + SpaceXAI are training a larger model from scratch with 10× more compute.

Leaks: “Vega”, 6 variants, Fast→Medium→High→XHigh reasoning.

Claims it beats GPT-5.6 Sol/Opus 5 are still unverified. 👀