Serial Resigner

Skip the professional summary entirely at this stage. Every graduate summary reads the same way because everyone is describing potential rather than history, and it burns the most valuable space on the page. Start with education, then projects, then work. You can add a summary in three years when you have something to summarise.

Do the call. Not forever - do it for the next 20 people and write down what happens. Within about three of them you'll know whether the blocker is generating the key, the permissions screen on the other service, or that they signed up out of curiosity and were never going to connect anything. Those are three different problems and the demo workspace only fixes one.

I did this and found the real issue was that the other service's admin screen needs a role most of my users don't have, so they'd have had to ask their own IT. No onboarding copy fixes that. A copy-paste 'ask your admin' email template does.

The bulk sender requirements kick in at 5,000 messages a day, so at 30 none of this applies to you. No DMARC needed, just send from a normal Gmail account and stop overthinking it.

Four screens from thirty is a good rate and it is real evidence about your actual market, whereas the score is a number generated by a company that would like to sell you a subscription. Small sample, yes, but it is a sample of the thing you care about rather than a proxy for it.

It is the DMARC record, but only because you never published DKIM. You went from "unaligned and nobody was checking" to "unaligned and you explicitly asked every mailbox provider to reject it".

Walk the header you posted:

• dkim=none because the CNAMEs your ESP handed you were never added to DNS.
• Return-Path is on the ESP's bounce domain, so SPF passes but does not align with mydomain.dev.
• DMARC needs SPF or DKIM to pass and align. Neither does. You published p=reject. Gmail did what you asked.

Order of operations:

1. Add the DKIM CNAMEs, wait for the ESP to show verified, send yourself one and confirm the header says dkim=pass header.d=mydomain.dev.
2. Set a custom bounce/return-path subdomain, usually mail.mydomain.dev, so SPF aligns too.
3. Drop DMARC back to p=none with a rua= address, read a fortnight of reports, then p=quarantine, then p=reject once the reports are boring.

Your open rate never fell. Your delivery location moved.

Freeze them, it kills the smell and you never have to wash them.

2,140 is small enough that formal warm-up is overkill, but do not send the first clean issue to the whole list. Take the 400 people who opened the last three issues, send to them, and let the engagement land first. Then everyone else the following week. The domain has no history, so the first signals it gets should be your best ones.

The push-it-to-a-call advice comes from a world where you can read a recruiter's reaction and adjust in real time, and some people are genuinely good at that. Most of us are not, and on a call you will get flustered and say a number lower than the one you wrote down. In writing you get to think for ten minutes and check your comparables.

Same experience with the readout. Mine spiked hardest when I made toast.

Not your situation but related and it caught me out badly: the very first payout on a new account can take considerably longer than the ongoing schedule, on the order of a week or two after the first successful charge. I had budgeted around a two day delay, told myself the money was late, and spent an evening convinced something was broken. It was just the first one.

I self-diagnosed from short videos and spent about nine months and a fair amount of money on planners, timers and productivity systems, all of which worked for about a fortnight each. When I finally got in front of a psychiatrist it was depression sitting on top of burnout, and the thing that helped was neither a planner nor a stimulant. The cost of guessing wasn't just the money, it was the nine months I spent being annoyed at myself for failing at systems that were never aimed at my actual problem.

If you are tempted by magic links, know what you inherit. They fold login and verification into one click, which is lovely, and they turn every email delivery problem into a login problem, which is not.

The specific one that will bite you: corporate mail security scanners follow links in incoming mail to check them, which pre-clicks your one-time login link and burns it. Then your best customer emails you saying the link is expired and they never even opened it. Allow the token to be used more than once inside a short window, or detect the scanner pattern, or you will get this ticket from exactly the accounts you least want to annoy.

Volume wins for the first screen and tailoring wins after it, which is why the comparison feels muddy. Nobody at the resume-reading stage cares that you used their words. The person who cares is the hiring manager reading it before the call, and by then you want the top of the page to look like the job they wrote. Spend the tailoring effort on the applications that reach a human.

Channel version, no contest. A goroutine owning the state is the idiomatic answer and the performance difference is theoretical at your volume.

The mistake I made was paying myself nothing meaningful for four years and calling it commitment. On paper the shop was profitable. When I finally priced my own hours in at what I'd have to pay someone to do them, three of our five services were losing money on every single job, and I'd been subsidising them with my own unpaid labour and calling it a business. Put a wage in the accounts even if you can't fully draw it, otherwise your P&L is a story about a person who works for free.

Most of that quote is labour and most of the labour is one job. Four bolts, a fuel line, a carburettor off, the bowl and the main jet cleaned, back on, and it is running. Do that once yourself and the mower is worth keeping for another five years, because fuel left over winter is the cause of about four out of five spring no starts.

Just use more lube, that's what it's for.

For the price of that quote I bought a cheap hygrometer for each room, a pin moisture meter and an infrared thermometer, and had change. The picture came together in a week: that corner sat markedly colder than the rest of the wall, humidity in the room climbed overnight, and the wardrobe against the cold wall was stopping any air moving across it. Pulling the wardrobe six inches off the wall and running the extractor properly did more than any report would have.

Just inflate it by 15k, everyone does and nobody checks.

Forty applications in six weeks for mid-level ops is a very small sample and I would not conclude anything from it yet. When I tracked mine properly the ratio was roughly one screen per twenty-five applications before I changed anything, and one per twelve afterwards, and I only knew that because I kept a spreadsheet with date, role, source and outcome. Without that you are guessing about the document when the variable might be the market you are aiming at.

Show output first, ask for credentials second. Every product with a connect-step graveyard has this backwards.

I added a three item checklist and learned something odd - people completed step one and step three and skipped step two, which was the integration, which was the entire point. Turns out they were doing the easy ones for the satisfaction of the ticks. Moving the integration to the end, after they had already got a small result out of the product, roughly doubled the number who did it. Same three steps, different order.

Weather is the one I'd add to the boring list. Pressure drops before a front do it for me reliably enough that I now check the forecast the way other people check a diary, no diet has ever done anything comparable.

Public transport is always fine, just take the night bus.

I think the weeknight versus weekend framing is the wrong axis. Four scattered ninety minute evenings gave me four context reloads and very little real work, whereas one uninterrupted five hour Saturday gave me the only architectural progress I ever made. Now I do one long block at the weekend for anything that needs thinking and use weeknights only for small, defined, already decided tasks. Same total hours, considerably more shipped.