Building @curn_io, AI-native Advisory &. Platform. Just starting on X. If you're a fan of AI and Strategy let's chat. Ex-strategy consultant → AI engineer.
@emollick Unfortunately, not very optimistic on this! Feel there are far too many vested interests in maintaining advantage on one side and maintaining momentum and acceleration on the other. Hope I'm wrong
Before I care how many steps an agent can handle, I want to know how painful the undo is.
Can we identify everything it touched, stop the next run and restore the previous state?
If a mistake is hard to reverse, make the first test smaller until failure is cheap and visible.
Beside any AI output that changes a real decision, save four things:
• the evidence it used
• the workflow version
• the human correction
• the accepted result
That is enough to ask: why did we trust this, and what would we change next time?
The first question in a lot of AI projects is: which tool should we use?
I think that is one question too early. Start with the part of the work people already find painful or keep checking twice. Otherwise, you are shopping for software and hoping a use case appears.
The boring workflow is often the interesting market. If it happens every week, costs real time and contains more exceptions than generic software can handle, it may be newly buildable. The dullness keeps competitors away while the exceptions give the product somewhere to learn.
AI businesses may be splitting into factories and instruments. Factories sell finished output, while instruments help somebody exercise judgement they already possess. The first competes on cost and speed; the second can become part of how a profession thinks.
The rate of innovation is speeding up.
But the rate of adoption?
It’s not just not keeping up, it’s slowing down.
I won’t even bring statistics in yet, just ask yourself the question: how overwhelmed have you been at all this new stuff?
I think we're teaching people to use AI in the wrong order.
Give them work they know and let AI flag an issue. They check the evidence, decide if it's right, fix it and run the check again.
Can they explain why? That's the skill worth rewarding, not how often they use the tool.
Wild observation this morning
Two apps I love and use every day, Granola and Littlebird. Littlebird recently came out with the latest meetings note update so just wanted to test it. (see below) Obviously, not the only company to follow Granola but this is probably one of the more brazen design-wise.
If digital products are now so easily copied, what does this mean for truly sustained competitive advantage?
Quick tip: use Gemini 3.5 flash for anything practical when engineering / vibe-coding. It’s extremely quick and seems they’re specialising the model on general tool creation and calling.
I asked it to create a mac computer use script for antigravity so it could interact with my mac like codex or claude and it did so in a few prompts.
If anyone’s using Gemini for any other great use cases please tell me in the comments, would love to try new.
Excited for this - something @Thomas_Beckley
and I have been working on for the past few months. What if you never had to repeat yourself again - to humans or your agents?
Every team has someone who's figured out the best way to use AI.
Right now, that knowledge is trapped in their head, or scattered across Slack, chats and docs.
Introducing @ascend_platform - the AI logbook that turns your team's best thinking into replayable playbooks and skills anyone can run, collaborate and build on.
One person's breakthrough becomes the whole team's starting point. Never repeat yourself again - to people or agents. Build your company brain.
Learn more at https://t.co/IlgxA3Welv
The most useful person on your team this year is the one who figured out how to actually use Claude. Or Cursor. Or Cowork. You can probably name them right now.
Every team has someone who's figured out the best way to use AI.
Right now, that knowledge is trapped in their head, or scattered across Slack, chats and docs.
Introducing @ascend_platform - the AI logbook that turns your team's best thinking into replayable playbooks and skills anyone can run, collaborate and build on.
One person's breakthrough becomes the whole team's starting point. Never repeat yourself again - to people or agents. Build your company brain.
Learn more at https://t.co/IlgxA3Welv
They didn't write a doc. They just got good at it. Quietly. Their best workflows live in their head, their DMs, and a few browser tabs. The rest of the team plays catch up.