# now
what i'm focused on right this minute. updated when the focus shifts. see the date at the bottom for the last revision.
## right now
main focus: using ai blueprint as the workflow layer for real builds, not just as a repo template. the goal is to make planning, implementation, checking, status, and adoption feel repeatable enough that every product block starts with the same rails.
product-wise, vidpipe is back near the top of the list. the public pages are crawlable now, the model routing is cleaner, and the next useful push is publishing: wordpress first, then the rest only if the workflow earns it.
multiprompter is the smaller build i keep wanting for recording: one script, multiple synced windows, less awkward script-reading during videos. still beta, but it is the right kind of personal tool.
the automation fleet also got rebuilt around a cleaner orchestrator and separate work lanes. that matters less as a public product and more as the operating layer under the rest of the work. i wrote the current shape up in my agent stack changed again.
## currently building
turning the blueprint into the default workflow layer: clearer gates, better resumes, status/check/adopt flows, and real use on active builds.
back in the product lane. seo prerendering is in, publishing destinations are the next serious feature, and the bundle needs a cleanup pass.
mac teleprompter experiment. one script driving multiple synced viewer windows, built for recording setups where the script has to follow me.
still the private ops surface for the vault. current arc is making the automation fleet easier to see, trust, and hand work to.
using the site as the public trail for what shipped: articles, devlogs, project pages, and the small tool catalog.
background publication lane. keep the quality bar, but it is not the main sprint while product work gets the deeper block.
## what i'm thinking about
- how much of product work can be turned into a calm blueprint workflow without making it feel like ceremony. the sweet spot is just enough structure that the ai does not drift
- whether vidpipe is a product problem or a distribution problem. the crawlable-site work fixed the technical floor; now the question is whether the publishing flow is useful enough to push harder
- personal tools as unfair advantages. multiprompter, linesmith, typesmith, and the agent workflow are all small things that make recording and building less annoying
- stackmaven is still interesting, but it is lower in the stack right now: maintain the editorial lane, avoid filler, and let the automation catch the high-signal stuff while the product work gets the deeper focus
## what i'm not doing
- no live coding sessions or live streams right now
- not chasing every shiny model release. only swap when there's a real reason
- not building a course platform. that's not what this site is for
- no new sponsor commitments beyond existing ones
// last updated 2026-06-29. inspired by /now pages.