Fleshed out plan for what we need to do to migrate fulfillment to an external 3PL vendor
We’re currently roughly manufacturing at the same rate we’re selling, but we need to get a surplus in order to cover the transition time
We had to figure out exactly how much of a head start we’d need and what else we’d need to do to ship the products to the 3PL
Worked with 3PL vendor on translating our workflows to them
Revised fulfillment processes to buy us time to switch to the 3PL
Cut out a time-consuming validation step that has never caught any issues
Deferred upgrade requests until mid-April
Met with attorneys for R&D tax credit
So far, not so painful
Interviewed and hired a new accountant
They’re a mid-sized firm and $5k/yr, which is a huge jump from my usual $300-500/yr
I’ve had such a frustrating experience with small, local accountants, and I hate researching tax questions myself, that I’m ready to pay to outsource it more intensely
Two calls with people with electronics manufacturing experience
Two 1:1s with local staff
Handled a request from a large customer who wanted to sponsor a new feature
I quoted $15k to do an initial investigation, and they stopped responding
I don’t have bandwidth to project manage a major new feature for a big company, so if I was going to make time for it, they’d have to pay a lot
Team lunch with local staff
Figured out bookkeeping for overseas business expenses I paid in cash
I wasn’t sure how to do double-entry bookkeeping in a foreign currency, but my reading of this post was that I should convert each transaction to the equivalent dollar value
It feels a little dirty because I’m failing to capture information (the real EUR amount, the exchange rate I’m using), but I’m preserving them in a comment
It was only a handful of transactions, so if I had to do this more regularly, I’d find a better solution.
pi-expand is a tool we’ve just started using internally to flash microSDs to the exact right size
Normally, Pi images flash only a portion of a microSD disk, and then on first boot, they expand to occupy the full disk
The problem is that the expansion adds 10-15 seconds to first boot, and if the user cuts power during their process, they likely cause an unbootable device
Since we know the size of our disks in advance, we can flash the full disk and save the user from the resize step
Investigated a bug related to pi-expand
One member of the fulfillment team reported that after we started flashing microSDs with pi-expand, audio streaming stopped working
This was bad news as we had already flashed several microSDs and wouldn’t be able to track which were flashed in the old way vs. new way, potentially forcing us to unwind 2-5 hours of work
I couldn’t reproduce it
It turned out to be bad instructions. I forgot that audio is off by default and didn’t explain how to turn it on.
We hadn’t run into it before because our standard testing microSD was pre-configured for audio
Fixed a bug related to email address string casing in license checks
According to the spec, two emails can vary by case, but no email server works that way, so when a user enters their email address in a different case than our records, we should ignore casing
I probably should just shut this down, but it was fun to see what it would be like reimplementing something I did five years ago with the knowledge I have today
Original version was Python 2.7 AppEngine + Cloud Datastore, but Python 2.7 AppEngine is going away in a few months
New version is pure Go with no datastore
The downside of no datastore is that everyone’s quota resets on every deploy (i.e., they get more free requests), but I don’t care that much, and it’s not worth the engineering effort
Reimplementing it was fun, but I don’t think I was that much faster than the original
Original probably took me 10-20 hours of dev time based on commit timing
New one took 5-6 hours
Misc
Presented at NERD Summit
Video is up but the audio is out of sync, so my slides give away all my jokes before I do
