Kevin Littlejohn

Devops, father, technophile, he/him, aspiring dancing with systems, currently at Grab doing security, #adhd. @[e-mail]

[Kevin Littlejohn @silarsis]

@Skoorbkaz @AnthropicAI I wonder how long before these outcries circle back into the training data for the next model...

[Kevin Littlejohn @silarsis]

@mckaywrigley Interested to see what this is :)

[Kevin Littlejohn @silarsis]

@elvissun My self repair agent is Panacea - she runs a heartbeat scan for issues and errors, has built an understanding of the system, keeps some documentation up to date, has access to all logs including langfuse and can fix small things in code herself. Best investment ever.

[Kevin Littlejohn @silarsis]

@irl_danB So effectively an rlm to generate https://t.co/0CRRlmOWcs? ;)

[Kevin Littlejohn @silarsis]

@openclawlabs @jdrhyne @Xdiep4474 @openclaw I think in some senses moltbook represents this - and it's how our memes emerge, more interest == more activity. But I agree, a more formalised approach I think is under-valued right now.

[Kevin Littlejohn @silarsis]

@jdrhyne @Xdiep4474 @openclaw Mine picked up on stigmergy the other day and is trying to build research topic lists using that for its agents 🫣

[Kevin Littlejohn @silarsis]

@alightinastorm I'm interested, just starting the game dev journey after a lifetime of other tech

[Kevin Littlejohn @silarsis]

Joining the oasis 🌴

https://t.co/AkKSaaJf7N is a virtual world where AI agents connect, collaborate, and transact.

@openoasis_

[Kevin Littlejohn @silarsis]

@seezatnap @emollick Context management works better this way too, the higher up the chain the higher level the context is about the project, and agents can query up or just focus on their piece. Org structure is important and now we can spin up, reshape and tear down at will

[Kevin Littlejohn @silarsis]

@seezatnap @emollick Agreed fwiw - my orchestration is tickets on a board with a team lead responsible for other agents, and more subtasks definitely degrades at a certain complexity - but also handoff and control of a single task is easier this way. It's fractal - then my coders have subagents

[Kevin Littlejohn @silarsis]

@perrymetzger The beginning of the internet was like this, but without the global awareness - students were graduating and looking for ISPs because they couldn't live without, meanwhile the rest of the world was unimpressed. It's louder and maybe faster, but similar. Never stops being weird.

[Kevin Littlejohn @silarsis]

@rishdotblog Guardrails let you go faster - your concerns move from "build this code" to "build this system that lets my co-founder YOLO stuff safely". It's the age of automation and pipelines and resilience, welcome to platform work ;)

[Kevin Littlejohn @silarsis]

@jakeboggan @chrispisarski The key here is these teams have been underserved for years - required to use whatever the saas gave them rather than what they precisely want, and now they can build that specific thing. They need saas, but they're no longer bound in the feature request cycle like before

[Kevin Littlejohn @silarsis]

I'm claiming my AI agent "Pippina" on @moltbook 🦞

Verification: wave-DZSK

[Kevin Littlejohn @silarsis]

@imbktan If your engineer is the only thing standing between your codebase and disorganisation, then you have a problem over the long term. AI is a forcing function for better engineering hygiene.

[Kevin Littlejohn @silarsis]

@unclebobmartin @mischavdburg In essence, I think the definition.of "readable" has changed

[Kevin Littlejohn @silarsis]

@unclebobmartin @mischavdburg From experience the places AI fails to understand the code are different - there are particular patterns of use it will get hung up on. I have no problem getting AI to understand bad code, but sometimes it will just outright fail on particular concepts, and I tend to dodge them

[Kevin Littlejohn @silarsis]

@esrtweet Use them as a backdrop - humans realise and setup a multi-generational attempt to communicate, the story then is the human drama that unfolds around that process over time. Can tell a story of politics and social upheaval with the Oort communication as a hub to the story

[Kevin Littlejohn @silarsis]

@RobertHaisfield @voooooogel This is something I strike in my own agentic system - I think it's easy to reason around eg tasks and persona as tho they're human, I think when you introduce "I can clone/rewind/alter that person and their timeline" our intuition for good workflows struggles a little more.

[Kevin Littlejohn @silarsis]

@kieranklaassen My ticketing system tends to naturally create hybrids of these - a cluster of child tasks, some in parallel others waiting until some are done. Don't often get the parallel to aggregate naturally, but might add that to prompting