350 Applications Got Me Three Interviews. 75 Got Me Two.

I applied to 450 jobs in 2026, three different ways. Here is what each way got me, and the one change that tripled my interview rate.

I didn't just lose a job. The jobs stopped existing.

I got fired because the industry I worked in fell off a cliff. Not my company specifically — the whole industry. Our company did a round of layoffs and I was in it, along with a lot of people I worked with every day.

I was in developer relations in crypto. If you haven't worked in that world, the title needs explaining: a dev rel is the person who goes from hackathon to hackathon, maintains the developer resources, and presents the product and the tooling to other engineers. Something between a marketer, a salesperson and an engineer. Mostly an engineer, but a sales engineer in the sense that you are selling to people who can read your code and will judge you for it.

I was senior dev rel and then got promoted to Developer Experience Director, which in practice meant leading two senior developers. So my CV is not weak. I am not a fresh graduate wondering why nobody is calling back.

I still couldn't get anyone to call back.

Everyone tells you to use your network. Mine was the industry that died.

Every piece of job-hunting advice starts with referrals, and the advice is correct. A referral is worth more than anything else you can do. I know this because I was on the hiring side too — we would get an enormous number of applications for one role, and a referral was the thing that moved a name from the pile to the top of it.

A referral is not a guarantee. One friend I personally referred didn't make it through the interviews. Another friend got referred by someone already inside and went through every round. But in both cases they got into the process, which is the part most people never reach.

The data is more lopsided than most people realise. Glassdoor's analysis of interview sources found that employee referrals produce the highest chance of a successful job match of any channel, while online applications with no personal connection to the employer carry an 11 to 15% lower probability of an accepted offer. And yet referrals account for only about 10% of reported interviews. Online applications account for 42%. The channel that works best is the one almost nobody gets to use.

Here is my problem. My network is crypto. The people who would refer me are the people who were laid off next to me, or who are sitting at companies that have stopped hiring because the funding stopped. The industry is in a squeeze, and a referral into a company with a hiring freeze is worth exactly nothing.

This is not a feeling, it's a measurable collapse. Tiger Research's H1 2026 analysis of the global crypto hiring market found that new postings on major crypto job boards fell roughly 80% year-over-year in January 2026, and that the market still has not recovered to its 2022 peak. A recruiter tracking crypto-only boards counted 85 to 90 unique new roles in the first two weeks of January. Across all of crypto. Meanwhile Coinbase, Gemini, Crypto.com, Kraken, Algorand and Messari all announced cuts in the first half of 2026.

So when people tell me to work my network, what they are describing is a network of people who are all applying to the same 85 jobs I am.

So the one lever everybody recommends was the one lever I didn't have.

And then AI changed what they were hiring for

The second thing that happened is that the job I used to do changed shape.

My old work was hands-on: reading code, writing code, reading other people's code, talking to people about code, showing people how the code works. A lot of code.

Now anybody can produce code at a junior or mid level. So the demand for juniors, for mid-level engineers, for people who broadly know what software is, has thinned out. What companies want is seniors — people who have written enough software to look at something and know whether it will hold.

This one has hard numbers behind it. A Stanford Digital Economy Lab study, Canaries in the Coal Mine?, analysed ADP payroll records and found a 13% relative decline in employment for workers aged 22 to 25 in the occupations most exposed to generative AI. For software developers in that age band, employment was down nearly 20% from its 2022 peak by July 2025. Workers aged 26 to 55 in the same occupations stayed stable or grew.

Read that again, because it is the whole story in one sentence: the jobs didn't disappear. The entry rungs did.

I don't think software developers are going away. I think the opposite. Production of software went to a hyperscale, which means what is actually scarce now is testing, quality assurance, and people who can tell good output from plausible output. The need didn't disappear. It moved.

The industry agrees with me on this, loudly. Qodo's 2026 State of AI Code Quality report, based on a survey of 500 US developers and engineering leaders, found that both groups independently named the same top bottleneck in software delivery: reviewing and validating AI-generated code. A SmartBear survey of 273 software leaders published in May 2026 found that 70% say application quality has already degraded as AI accelerated development, with 60% reporting quality problems because code creation outran testing capacity.

The bottleneck moved from writing code to trusting it. That is a job. It is just not the job my CV described.

But if the shape of the job moved and your CV still describes the old shape, you are applying into a gap. You have to rebrand. That turned out to matter more than I expected, and I'll come back to it, because it is the single thing that changed my results.

Even inside my own dying industry the shift is visible in the postings themselves. Tiger Research found that the share of crypto job listings mentioning AI skills rose from 23% in early 2025 to 53.1% by March 2026. Half the remaining jobs in crypto now want something my old title doesn't say.

The warm intro works. Count what it costs.

I got my last job by walking up to the CEO at a conference. I liked what the company was saying, I asked him directly whether they were hiring, and that was that. It worked. It is a far stronger signal than an application form, and I would recommend it to anyone.

Now count what it costs. To do that, you have to be physically at the conference. You have to travel there, which is a week or two of your life. Once you're there, if you are very good, you will connect properly with one, two, maybe three companies.

And connecting properly is not a handshake. To get that warm feeling from them you sit through their talks, you talk to several of their engineers, you show real interest, maybe you join their hackathon and build something on their stack. That is a serious investment of your time in one company.

After all of that, you still don't know whether you'd pass their interview. It might be the only company you're interviewing with, and you can fail it for reasons that have nothing to do with how well you worked the room.

So: the highest-conversion channel, with the lowest volume and the longest cycle. If you can travel, do it. But if you can't get on a plane every month, or you have a kid, or you simply don't have the runway, there is one channel left. You apply. A lot.

25 applications by hand. One interview.

My husband's advice was blunt: apply to as many companies as you possibly can, line up as many interviews as you can, and let's see how many people respond. I resisted it for weeks. I had exactly one interview, from a recruiter who found me on LinkedIn, and no offer.

I was procrastinating, and the procrastination had a respectable disguise. I thought I needed more LinkedIn followers. A few more projects so my GitHub looked better. More talks. A better-looking YouTube channel. A rewritten CV.

Here is what I now believe: start applying with the least preparation you can get away with. Whatever CV and references you have right now are probably enough to find out what's wrong with them. You cannot guess which part is holding you back. Recruiters will tell you, and their feedback is specific in a way your self-doubt is not. Polishing before you have any feedback is just a nicer-feeling way of not applying.

Eventually I got annoyed enough to prove him wrong. I went company by company. I applied to Kraken — opened their careers page and applied to maybe 10 or 15 roles there. Then OpenZeppelin. Then a few other crypto companies I knew and respected.

About 25 applications, hand-written. The whole thing took the better part of a day, maybe two. And looking back, plenty of them were for roles my CV didn't actually match — I was applying to whatever was open at a company I recognised, not to work I was right for. One company invited me to an interview.

It was a Slovenian company, small team of young developers and marketers, salary not great, and at 29 I would likely have been the oldest person there. My husband told me to go to the interview anyway. He was right about that too.

And I want to name the part nobody writes about. Applying by hand is humiliating. You are putting yourself in front of stranger after stranger to be judged, and most of the time the answer is an automated no, or nothing at all. I know exactly how that pile looks from the other side, because I used to be the one reading it. Knowing doesn't make it feel better.

And the arithmetic is brutal in a way that should make you feel better about your own silence-filled inbox. Greenhouse, which hosts around 175,000 live jobs, now averages roughly 254 applicants per opening, and applications per recruiter have climbed 412%. Recruiters describe spending between 30 seconds and two minutes on each CV, while handling 15 to 20 open roles at once. Popular postings collect 150 to 200 applications in the first 24 hours.

Twenty-five hand-written applications, in that environment, is not a job search. It's a rounding error.

350 applications, three interviews, and an empty meeting room

If volume is the only lever left, I did not want to pull it by hand. So I went looking for something that would apply for me.

First I tried to build the pipeline myself with Claude Code. That was ineffective in a way I should have predicted. It burned credits finding jobs, burned more credits applying, and then wanted me to manually review everything anyway. I automated my way back to doing it by hand.

Then I searched for auto-apply tools. The first page is all sponsored results, and several of them advertise automatic applications and then simply don't apply automatically. You sign up, you pay, you still fill in forms.

I eventually found one that genuinely worked, and I used it to send 350 applications.

Three hundred and fifty. Out of those, three interview invitations, including one at an AI-focused company and one in crypto. By volume standards that is the system working.

Then I sat in a video call for a second-round interview and nobody came.

The tool gives you a mailbox where recruiter replies are supposed to land. The replies were landing, technically, but the inbox was structured so badly that I never saw them. I had written a personal email confirming I would be there. They had replied. I never saw the reply, they never saw a confirmation they could read, and so they assumed I wasn't showing up.

I nearly lost one of the only real opportunities I had in months, because of an inbox.

That was the moment the project started. Not out of ambition — out of being furious in a quiet room by myself.

Why most of those 350 were wasted

The inbox was the dramatic failure. The quiet failure was worse, and it is the one that actually taught me something.

Most of those 350 applications went to jobs that had nothing to do with me.

The reason is how the matching works. You pick your roles from a generic list: Software Engineer, Product Manager, that kind of thing. Then the tool fires your CV at everything carrying that label, anywhere. I tried narrowing my description. I tried widening it. Nothing moved. At some point it looked like the system simply had no jobs for me at all, which for a person with a decent CV is an absurd thing to be told by software.

And I was not the only person doing this. LinkedIn now processes around 11,000 job applications per minute, a 45% jump in a single year, driven largely by AI tools letting people fire off hundreds of applications they never read. The number of jobs did not rise 45%.

The result is exactly what you would expect. In a LinkedIn survey, around 70% of hirers said fewer than half the applications they receive meet the criteria for the role, and one recruiter put it plainly: the wrong people are now applying en masse to the wrong roles, because AI made it free to do so. One headhunter described a single popular listing collecting over a thousand applications across a weekend.

So the volume strategy has a trap in it. If you use a blunt tool, you are not beating the flood. You are the flood — and you are training recruiters to distrust exactly the kind of application you are sending.

My husband and I had the same conversation over and over: are these even being sent? Are they arriving? Do I look like a scammer to these people?

That last question wasn't paranoid. The tool had a feature that rewrites your CV for each posting, and the output was bad — it misread what I'd given it, restructured nothing, and produced something that read like a fake person applying. I turned the feature off completely. I would rather send a plain honest CV than a generated one that makes a recruiter suspicious in the first five seconds.

So: enormous volume, generic targeting, mangled CVs, and replies I couldn't see. Four problems that each have a fix.

So I built NightApply

I build developer tools for a living. At some point it became ridiculous to keep complaining about software instead of writing it.

NightApply tracks around 177,000 open positions from company career pages and job boards, and drops the ones that close. That is the easy part — scale is a solved problem, and it is not what was broken.

What was broken is targeting, and the fix starts with your CV. NightApply reads it, pulls out the keywords you actually use, and tells you which roles your experience matches. Not categories. Specific titles.

When I ran my own CV through it, it suggested Founding Engineer, Blockchain Engineer, SDK Engineer, Smart Contract Engineer — and a lot more I hadn't considered. This is where I learned something slightly painful: almost nobody is hiring a "Developer Experience Director." That was my title. It is not a job posting. I had been searching for the title I used to have instead of the work I can do.

The list I apply against now looks like this:

• Developer Experience Engineer, Staff Developer Experience Engineer, Developer Platform Engineer
• Developer Relations, Developer Relations Lead, Developer Relations Manager, Developer Relations Engineer, Developer Advocate Lead
• Harness Engineer, Agent Experience, Agentic AI, AI Product
• Technical Writer, Technical Content Lead, Developer Tool Engineer
• Engineering Manager, Solutions Engineer, Product Manager

That is the rebrand I mentioned at the start, and it is not a story you tell about yourself. It is a list of search terms. Every one of those titles gets matched against the open roles in the scanner.

Then, before anything is sent, your CV is scored against the specific posting — so you see where you match and where you don't, on that job, before it goes out.

The rest is the things that failed me:

• A mailbox that works. Every recruiter reply lands in one place, readable, so you never again sit alone in a meeting that someone did show up to.
• First in line. It applies the night a matching role is posted, before the first-day pile of 150 to 200 applications builds up.
• Auto mode and manual mode. Let it send, or approve each one yourself. A browser extension handles the applications that need you in the loop.
• No fake CVs. Keep several CVs, one for each kind of role, and NightApply picks the one that fits each posting. Rewriting a CV per job is a different thing: that ships only when it produces something I would be willing to send under my own name.

Where I am right now

I am 75 applications in with my own software. Two interview invitations so far.

That is the whole argument of this post, so here are all three attempts side by side:

Per application, the matched batch got interviews about three times as often as the generic one: 2 in 75, against 3 in 350. Same CV. Same person. Same miserable market. Applying by hand did fine per application, but 25 of them ate two days, and there is no version of my life where I do that 450 times.

I'm not going to pretend this ends with a job offer, because it hasn't yet. One of NightApply's two interviews I handled badly. The other is next week. Across all 450 applications, that is six interviews. My husband reckons it takes about ten to land something real. He has been right about every other part of this, so I assume he's right about that too.

What I can tell you is what changed. I stopped waiting for a network that no longer exists. I stopped polishing a CV I had no feedback on. I stopped applying under a job title nobody is hiring for. And I stopped losing interviews to an inbox.

If you are in a dead industry, or a dying role, or just tired of refreshing a careers page at midnight, that is what NightApply is for. It's what I built because I needed it. I'm still using it on myself every day, which is the only product guarantee I actually trust.

Start tonight: it takes about ten minutes to set up

Sources

Figures as of October 2026.

All application counts, interview counts and outcomes in this post are my own, from my own job search.