The advice to "apply to more things" was written for a market that no longer exists. Ashby, which makes hiring software and can therefore see the whole funnel, analysed over 109 million applications and 247,000 jobs from January 2021 through March 2026. Their finding, in their own words: "Candidates are roughly 50% less likely to receive an interview today than they were five years ago." Applications per hire went past 300, and the share of applications that turn into an interview fell from 7% to 8% in 2021 down to somewhere between 3.6% and 4.7% today.
Then look at what happens when someone knows your name. In a separate Ashby analysis of 38 million applications across 93,000 jobs, referrals were about 1% of all applications received, and yet "40% of referred candidates go from application to interview". By the end of 2024 the offer rate for referrals sat at 7.3%, against 0.2% for people who applied cold.
Do not run this out of scattered chats. Everything below works far better inside one Claude project, because the project carries your resume, your target list and your voice into every conversation without you re-pasting them.
You are helping me run an embedded job search. That means I am not mass applying. I am picking a small number of target companies, building genuine relationships inside them before roles open, and staying top of mind so that I am considered early when something appears.
Ground every recommendation in my actual resume and writing samples in this project. When you name a person, a team or a company fact, use web search and tell me the source and the date, and say plainly when you are not sure, because I am going to contact real humans and being wrong about them is worse than being slow. Never invent a job title, a mutual connection, or a quote.
Write anything I might send in my own voice, matched to my writing samples: short sentences, specific, warm, no corporate filler, no em dashes, and never a paragraph that opens by flattering someone. Assume the person reading it is busy and slightly sceptical. If a piece of outreach has no reason to exist other than that I want a job, tell me so and help me find a real reason instead.Fifteen is the number because it is small enough to be real. You can hold fifteen companies in your head, notice their news, and have a genuine opinion about them. Fifty target companies is just mass applying with extra steps.
Using my resume in this project, build me a target list of 15 companies for an embedded job search. I am based in [your city] and open to [remote / hybrid / on-site], and the work I want to be doing is [describe the role in plain words, not a job title].
Rank them by leverage, not by whether they are hiring right now. Leverage means the overlap between what they are visibly trying to do and something in my background that is genuinely uncommon, so for each company tell me the specific thing they are working on, the specific thing in my resume that meets it, and why that combination is rarer than it looks. Search the web for recent signals: funding, launches, expansion into a market, a new leader with a stated agenda, a public problem they are trying to solve.
Give me a mix: five companies where I am an obvious fit, five where I am a strong fit that they would not think to search for, and five that are earlier-stage or less obvious and where one conversation could matter disproportionately. Include a source link and a date for every signal you cite. At the end, flag any company you were not able to verify current information about, and tell me the three companies from the list you would drop first if I wanted to cut it to twelve.Not the recruiter, and usually not the CEO. The person you want is one or two levels above the role you would take: close enough to feel the problem, senior enough to create a headcount when they feel it badly.
For each of my top 8 target companies, map who I should actually get to know. For each company give me three names: the person who would most likely be my future manager, one peer already doing work adjacent to mine, and one person who talks publicly about the company's work, for example on LinkedIn, a podcast, a conference talk or an engineering or marketing blog.
For every person, use web search and give me: their current title, roughly how long they have been there, a link to the profile or post where you found them, and one specific, recent, verifiable thing they have said or shipped that I could genuinely respond to. That last part is the whole point, so if you cannot find something specific and real for a person, say so and suggest someone else rather than filling the gap with a generic description.
Then tell me, in one line per company, which of the three is the right first contact and why. Rank all 8 companies by how reachable their people actually are, because a warm door at a smaller company is worth more than a cold one at a famous name.The message that works is the one that would still make sense if you were not job hunting at all. You are giving something small, reacting to something specific, or asking a question only that person can answer. The job comes up later, or it never comes up and they think of you anyway.
Draft first-contact messages for the five people you ranked highest. Match my writing samples in this project so they sound like me, and keep each one under 90 words.
Each message must be built on the specific thing that person recently said or shipped, and must contain something of my own: a reaction they have not heard before, a relevant observation from my experience, or a question that is genuinely interesting to answer. None of them may ask for a job, a referral, a coffee chat, or "15 minutes of your time", because those all put the cost on them and give them nothing back. If the only thing I have to offer a particular person is enthusiasm, tell me the message is not worth sending yet and what I would need to learn or make first.
For each one, give me the message, the platform to send it on, and a one-line note on what a good reply would look like versus what a polite brush-off looks like, so I know how to read the response. Then write me one short alternative version of each that I could post publicly as a comment instead, since a thoughtful public comment is often a lower-friction first touch than a DM.Then the part almost everyone skips, which is what you say the second and third time.
For each of the five people, build me a three-touch follow-up ladder spread across the next eight weeks, assuming the first message got either a short friendly reply or no reply at all.
Each touch has to carry new value rather than repeating the ask, so give me concrete options: sharing something relevant I genuinely read, reacting to a company announcement, sending a short piece of work or analysis I made that touches their problem, or answering a question they posted publicly. Tell me what to do differently if they replied warmly versus if they went silent, and be honest about when to stop, because persistence stops being charming at a point and I want to know where that point is.
Write the actual draft text for touch two and touch three in my voice, leaving obvious blanks for the thing that will only exist later. Then tell me the single strongest touch on the whole ladder and why, so I know which one to put real effort into.An embedded search dies from forgetting, not from rejection. Fifteen companies and forty people is more than anyone holds in their head, and the whole strategy depends on the second and third touch happening at all.
Build me a simple tracker for this search as a table I can paste into Notion, Google Sheets or Excel, and give me the rows already filled in with my 15 companies and the people we mapped.
Columns: company, person, role, platform, first contact date, last touch date, what the last touch was, their response, next touch date, next touch idea, and a warmth rating from one to five. Then tell me the rules for maintaining it in plain language: how often I update it, what warmth ratings actually mean in behaviour rather than feeling, and what triggers moving a company down the list or off it entirely.
Finally, write me a short weekly review prompt I can paste back to you every Sunday with the updated table, so you can tell me who is overdue for a touch, which relationships are going cold, and which two people I should put real effort into this week. Keep that review prompt under 120 words so I will actually use it.Recall is not the same as contact. These are the four that make people think of you unprompted, roughly in order of effort and of how much they are worth.