For most of my last year, I thought that was the smartest question a person could ask.
If I kept choosing the sensible thing, the defensible thing, the thing that made sense when I explained it out loud to someone older than me, I’d end up somewhere good. That was the deal. Do the reasonable thing enough times and reasonable things happen to you.
And it worked, sort of. It got me here. But somewhere along the way I stopped being able to remember the last time I’d wanted something for a reason I couldn’t justify. Everything had a rationale. Everything was a step towards. Nothing was just interesting.
I’m bringing this up because I’ve been watching a lot of people lose their jobs this year, and I’ve noticed something about how they look for the next one.
Let’s get the depressing part out of the way, because pretending otherwise is insulting.
Layoffs have stopped being an event and started being a weather pattern.
People are getting cut in waves, quietly, in Slack messages at 9am, often from teams that were profitable. Entire functions are being restructured around tooling that didn’t exist eighteen months ago. Junior roles are thinner on the ground than they’ve been in a decade, which is a genuinely bad thing for an industry that needs people who don’t yet know what’s impossible.
Here’s what I want to say plainly, because it doesn’t get said enough:
Being laid off is not a review of your work. It is a review of a spreadsheet you were never shown.
The people I know who were cut this year were not the worst people at their companies. Some of them were the best. They were expensive, or they were in the wrong cost centre, or their manager lost a fight two levels up. That’s it. That’s the whole story.
The reason this matters for the rest of this issue is that a layoff quietly rewires how you search. You get scared, and scared people apply to everything. Three hundred applications, no replies, and a slow creeping suspicion that the problem is you.
The problem is not you. The method is broken.
When people ask “what’s the right job for me”, they usually mean one of two things. Either what job will make me safe, or what job will make me impressive.
Both are fine questions. I’ve asked them plenty. But they’re the wrong ones to lead with, because neither of them is actually answerable from where you’re standing. You can’t know if a job is safe. You definitely can’t know if it’ll be impressive in four years, given that the entire hierarchy of what counts as impressive in this industry got shuffled sometime around last March.
There’s a better question, and it’s embarrassingly simple.
What do you read about when nobody is making you?
Not what you’re good at.
Not what you studied.
What you find yourself opening tabs about at 11pm for no productive reason at all. Evals. Robotics. Payments infrastructure. Climate. Devtools. Fashion supply chains. Whatever it is.
That’s your stream.
And the reason it matters isn’t spiritual, it’s tactical: you already have context in it. You already know the names, the arguments, the annoying unsolved bit that everyone complains about.
Context is the thing that makes a cold email land, and you cannot fake it in a week.
This is the part you can do this week. It is slow, low-volume, and it works far better than the alternative.
1. Pick the stream. The subject, not the job title. Be specific enough that it excludes things. “AI” is not a stream. “Agent memory” is a stream. “Voice for non-English markets” is a stream.
2. Find five to ten companies inside it that you actually like. The test is not funding, or headcount, or whether your parents have heard of them. The test is whether you’d use the product, or whether you’ve thought “huh, that’s interesting” about something they shipped.
If you can’t name a single thing you admire about a company, do not apply there. You will be miserable and they will be able to tell.
3. Ignore whether they’re hiring. I mean it. Check, obviously, and apply if there’s a fit. But job postings are a lagging indicator.
By the time a role is posted, it has been budgeted, argued over, and forwarded to six people’s friends. The interesting roles get created for people who show up before the posting exists.
4. Find the humans. Find one person doing the job you want at that company, and find the founder. Two names. Not twenty.
5. Bring a problem, not a résumé. This is the whole thing. This is the part everyone skips.
Spend two hours with the product. Actually use it. Find one specific thing that’s broken, missing, or underexploited. Their onboarding drops you into an empty state with no example. Their docs don’t have a quick start for the thing everyone actually wants to do first. Their competitors are all over a market they haven’t touched. Their pricing page doesn’t explain the thing that makes them better.
Then write four sentences.
Hi [name], I’ve been using [product] for a few weeks for [specific thing]. One thing I noticed: [the specific problem, described precisely]. I put together [a rough fix, a teardown, a mock, a doc, a small script] in case it’s useful. Either way, I’d love to talk about how you’re thinking about [the broader area].
No paragraph about your journey. No “I’m passionate about the space.” No attachment of a CV that they will not open. A specific observation, evidence you did the work, and one low-friction ask.
The reason this works is not that founders are charmed by initiative, although they are. It’s that you’ve just done the first day of the job in front of them.
Hiring is mostly risk reduction. You have removed the risk.
If you want the systems half of this, I wrote a full walkthrough a while back on setting up an AI-assisted job search: profile, tailoring, timing windows, the whole pipeline. It pairs well with everything above.
You will send twenty of these and hear back from two. That’s a great rate, and it will still feel bad on the days nobody replies.
It’s slower than mass applying. You cannot do this for two hundred companies.
That’s the point, but it is also a real cost when rent is due, so run the boring high-volume applications in parallel and don’t let anyone tell you that’s beneath you.
And it does not work if the problem statement is generic. “Your marketing could be better” is not a problem statement. It’s an insult with a subject line.
Some of the best things in my life started as choices I could not defend at the time. Writing things nobody asked for. Turning up to communities where I knew nobody. Reading papers I was under-qualified for and building with tools I half understood. None of it made sense on paper. All of it compounded.
So keep asking what makes sense.
It’s a good question and it’ll take you far. Just don’t let it be the only one you ask, because if you optimise hard enough for sensible, you will slowly stop making room for anything you love, and one day you’ll look up and not be able to remember when that happened.
Ask the sensible question. Then ask the other one.
Sierra — San Francisco HQ, with offices in New York, Atlanta, London, Singapore, Tokyo, Paris, Madrid, Munich, Sydney and Toronto.
Zero to $200 million in revenue in twenty-six months, which is the kind of number that used to take a decade.
Hiring for: Software Engineer, Agent (in nearly every office). Forward Deployed Infrastructure Engineer. Strategist, Agent Development. Product Manager, GTM Manager, etc.
Apply: sierra.ai/careers
Cribl — Remote-first. Currently listing roles for Remote US, Canada, Poland, Singapore and San Francisco.
Ranked third on Forbes’ Best Startup Employers list this year. They build telemetry infrastructure, which means they get paid more as everyone else spends more on AI, rather than fighting for the same customers.
Hiring for: Software Engineer, Cribl AI. Manager, Software Engineering (Cribl AI). Senior Software Engineers across Infrastructure, Lake, Product Platform, etc.
Apply: cribl.io/careers
ElevenLabs — Remote-first, team across 30+ countries, hubs in New York, London and Warsaw. Several roles can be done from anywhere.
The most realistic one on this list if you’re not in America and not planning to be. Lean teams, async by design, and they hire well outside engineering.
Hiring for: Forward Deployed Engineer. Product Designer. Compliance Engineer. IP Counsel. Mobile Content and Strategy Operations. Music GTM specialists.
Apply: elevenlabs.io/careers
Harvey — San Francisco, New York, Chicago, Dallas, London, Singapore, Sydney, plus remote. Hybrid at three days in office, with a remote option if you’re outside those cities.
They hire practising lawyers into AI product roles. Not engineers who learned law. Lawyers. On the Forbes AI 50 for two years running.
Hiring for: Legal Engineer. Legal Engineer, Product Specialist,etc
Apply: harvey.ai/careers
Replit — Foster City HQ, deliberately in-person. The SRE roles are remote across Europe.
Fifty million people use it. And their business development listing openly says vibe-coding experience is valued, which is the first sales job I’ve seen where being an enthusiastic amateur builder is an asset instead of a gap.
Hiring for: SRE (Remote Europe, including the UK, France, Ireland, Italy and the Netherlands). Senior Business Development Manager, etc.
Apply: replit.com/careers
One last thing. Look at those job titles again. Legal engineer. Agent strategist. Product specialist. Forward deployed engineer. Half of these didn’t exist three years ago, and most of them don’t require you to be an engineer.
That’s the whole argument of this issue, sitting right there in the postings.
To finding jobs and getting hired,
Shreya
No posts
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.