12 habits of a successful builder
what would you add?
STOP:
> trying to fit in. stand out.
> selling your time. invest it.
> working harder. work smarter.
> over consuming. create more.
> buying more. sell more.
> chasing clout. chase knowledge.
> obsessing over complex skill. build your soft skills and consistency.
> watching more. read more.
> saving money. invest more.
> watching other people’s bags. build your own business.
> working for people. hire people.
@markproduct lol that was my answer because I’m building a tool that’s translates the logs of what your agents are doing into natural language and categorises them so you can decide what you want to dig deeper into, so you can review the raw logs.
Don’t build with AI if you’re not comfortable waking up in the middle of the night to multiple auto-recharge credit emails 😅
Not for the faint hearted.
Lots of optimising to do today and Sonnet 5!!
Six months ago, if you wanted AI that could plan, use tools, and work independently at a high level, you paid premium prices for it.
Today Anthropic shipped Claude Sonnet 5. Available now for Free, Pro, Max, Team, and Enterprise users.
Sonnet 5 closes most of the distance to Opus 4.8 on agentic coding, multidisciplinary reasoning, and computer use.
I'll come back once I've spent a few days actually working with it.
https://t.co/GHOrUia9yH
@steipete I haven’t used the app but openclaw piers my agent infrastructure and I couldn’t be more grateful. Keep pushing, happy to help if you need help managing the team or getting things done
@Taniyatweets_ Outlook 😅
I lied, Gmail.
I’ve always wanted to try @Superhuman though, but Google have my entire ecosystem and paying for two email providers or a wrapper around Gmail feels meh
My agents and AI of choice are wired to check me straight away and never start with agreeing. Meaning I’m easily talked out of stupid decisions.
Today I asked Claude if I should invest in the Japanese Yen… because of course after I saw one tweet I was a forex expert.🤔😅
If you are someone who acts of hype, you need to wire your agents or AI of choice to always start with the opposing view of whatever you ask it unless it’s a simple task.
You will be talked out of stupid decisions.
If you are someone who acts of hype, you need to wire your agents or AI of choice to always start with the opposing view of whatever you ask it unless it’s a simple task.
You will be talked out of stupid decisions.
Vercel recently made every AI agent a folder you can open, read, and audit.
instructions.md - the system prompt
tools/ - what it can do
schedules/ - when it acts alone
This is the kind of AI governance we should have had from the start