Peter Kirkup

VP Innovation at Disguise. Proud dad. Spending my time building marble runs and playing with agents.

[Peter Kirkup @pkirkup]

@tymofii @Anthropic @coderabbitai Can you point to the docs on the chat processor hook? I can’t find it

[Peter Kirkup @pkirkup]

@anthropic @coderabbitai Can you both figure out how to suppress the <internal state> context that CodeRabbit injects with every comment - 40kb of context rot on every PR comment, no way to block it in Claude code when subscribed to the PR.

[Peter Kirkup @pkirkup]

@harjotsgill @Anthropic @coderabbitai Thank you! This is a killer for context at the moment!

[Peter Kirkup @pkirkup]

@_catwu Would love some slots on friendly EMEA times - popped a few thoughts here: https://t.co/VzPhbc6uiA

[Peter Kirkup @pkirkup]

@OmidMogasemi I’d love to chat but the slots are difficult - keep hitting a bug with sessions saying out of usage, even when topping up or extending usage limits. Crippling as the session is unusable after that. Business premium + extra usage. I’m account owner.

[Peter Kirkup @pkirkup]

@OmidMogasemi Other things that would close the gap for me - proper secret store (current secrets are basically unusable in business environment), and repo local .claude/skills skills not being supported / recognised on session launch.

[Peter Kirkup @pkirkup]

@OmidMogasemi I’d love to chat but the slots are difficult - keep hitting a bug with sessions saying out of usage, even when topping up or extending usage limits. Crippling as the session is unusable after that. Business premium + extra usage. I’m account owner.

[Peter Kirkup @pkirkup]

As world models emerge, the role of the media server is expanding.
Shows become real-time systems - evolving with timing and interaction.
See our latest proof of concept: https://t.co/9mt7YkLTvI @odysseyml @disguise_one #worldmodel #generativeai

[Peter Kirkup @pkirkup]

@mattshumer_ This feels like the real inflection point.

Once agents reliably handle execution, the structure of work itself starts to change. The whole “sit at your desk 9-to-5 doing tasks” model assumes humans must execute every step.

I wrote a piece exploring this:
https://t.co/DPjBoV5zph

[Peter Kirkup @pkirkup]

@karpathy This idea of “managing agents working in parallel” is fascinating.

If that becomes normal, work shifts from doing tasks to directing systems.

That probably breaks the traditional 9-to-5 structure entirely.

I explored that idea here:
https://t.co/DPjBoV5zph

[Peter Kirkup @pkirkup]

@levie I’ve been thinking about this a lot lately.
If agents start doing large chunks of the “task layer” of work, the structure of work itself probably changes. The 9-5 made sense when humans had to sit at the keyboard to execute every step.
I wrote a piece:
https://t.co/DPjBoV5zph

[Peter Kirkup @pkirkup]

The 9-to-5 workday was invented for factory lines.
But now we’re in a world where AI agents work 24/7.
Engineers are moving from writing code → orchestrating agents.
Launch task.
Review output.
Iterate.
So why are we still measured by hours at a desk?
https://t.co/DPjBoV67eP

[Peter Kirkup @pkirkup]

@ryan_tech_lab How do you think we will see specs and feature requests evolving - are we too unambitious, scoping out requests to the capabilities of historic dev resources?

[Peter Kirkup @pkirkup]

The scariest thing about AI agents isn’t autonomy.

It’s immediacy.

When every prompt pays out instantly, building becomes compulsion.

At what point does “shipping fast” become behavioral conditioning?

https://t.co/ol82EvJv4v

[Peter Kirkup @pkirkup]

@rohanpaul_ai It’s the dopamine! I wrote about this this week: https://t.co/sKXt8UMjW9

[Peter Kirkup @pkirkup]

@Adam_Wilbanks I wrote about exactly this - the dopamine hit is insane!
https://t.co/ol82EvJv4v

[Peter Kirkup @pkirkup]

@karpathy @theo @levelsio - what do you think, Agents Anonymous?

[Peter Kirkup @pkirkup]

@ryan_tech_lab Good question - if the goal is now easily scored, i find the goalposts keep moving. With some of my projects - I start with a goal - as soon as an agent is done, an urge to add more functionality, refine and iterate where previously the initial delivery would have sufficed.

[Peter Kirkup @pkirkup]

@XmaxAIOfficial Would love to give this a try!

[Peter Kirkup @pkirkup]

It’s here! Our FIRST official Disguise plugin for @bitfocusio Companion is now LIVE. 🚀
Go beyond simple triggers. Get real-time feedback from Designer directly on your Stream Deck buttons.
Available to all Companion users from today.
https://t.co/KQLBpcXv9s @disguise_one @elgato

[Peter Kirkup @pkirkup]

AI-generated video is getting serious. 🛠️The big news isn't just the visuals:
@LumaLabsAI
now exports HDR in EXR format. We tested in a full HDR pipeline in
@disguise_one
& it works flawlessly. AI tools are maturing for professional use. https://t.co/Z1AbQSJgwK #ProAV #AI #HDR