It is June, and the cold emails started two weeks ago. Every year around now my inbox fills up with two kinds of message. The first is from someone who just got a CS degree and wants to know how to land an engineering job. The second is from someone who wants to break into VC and thinks I have a door I can open. This year the two messages have started to sound the same. Both of them are asking, underneath the LinkedIn-polite phrasing, the same question. Is there still a way in?
Here's the deal. The answer is yes, but not through the door you are knocking on. That door got bricked up while you were in school.
The bottom rung got sawed off.
The numbers are ugly and I am not going to soften them. CS grad unemployment is sitting at 6.1% in 2026, roughly double philosophy majors, which is a sentence I never expected to type. Recent-grad unemployment hit 5.7% late last year, worse than any point in the 2008 crisis. Entry-level software roles are down about 30% year over year. Big tech hired more than 50% fewer fresh grads than it did three years ago, and only 7% of its new hires are recent grads at all.
Stanford put a clean number on the cause. Employment for software developers aged 22 to 25 has fallen nearly 20% since 2024. Not the 45-year-old staff engineers. The 22-year-olds. The exact cohort whose job was to write the boilerplate, draft the tests, and implement the well-specified ticket. That work was your training ground. It is now the agent's default output.
So yeah. The first rung of the ladder, the one every one of us used to climb, is the rung the model now does for free.
The other ladder never had a bottom rung.
Now the email from the aspiring VC. I have some bad news that is also, weirdly, good news. Venture was never going to hand you an entry-level execution job, because venture barely has entry-level execution jobs. There is no on-cycle recruiting. There is no grunt tier that trains you up. Analyst seats are scarce, base comp runs $80K to $130K, and almost every one of them gets filled through somebody's network, not a job board.
For a decade that was the cruelty of VC. You could not get in without already being in. But here's the thing. The trait that always got people the seat is the trait that now matters everywhere else too. Funds do not hire the person who can run a discounted cash flow in their sleep. They hire the person who has been in the room with founders, who has shipped a product, who holds an actual opinion about why a company is good. Operators get the seat. Finance kids get the polite rejection.
Which means the two emails in my inbox have the same answer.
Not just another AI newsletter. MavSource aggregates updates from all major AI newsletters, podcasts, company news, AI labs, and hundreds of other sources — then summarizes what matters, analyzes emerging trends, and adds founder commentary. One 5-minute daily email. Free.
Both worlds now pay for taste, not for typing.
Here's the thing nobody put on your syllabus. The execution floor is gone in both careers. The agent writes the boilerplate. The model summarizes the deck. What does not commoditize is judgment, and judgment is the whole game now.
If you want the engineering seat, stop trying to out-type the agent on volume, because you will lose, and your interviewer already knows you will lose. Go build one real thing that one real person uses, in public, and be able to explain every line of it even though an agent wrote half. That is proof of taste, and taste is the scarce input. Pick the edges the model is still bad at: distributed systems that actually fail in anger, security work that needs an adversary's brain, evals, hard domain depth. The generalist who knows Python is the commodity. The person who can decompose a vague problem and direct three agents at it is the hire.
If you want the VC seat, the move is identical. Be in the room. Build something, even a small thing that flops, so you have lived the founder's side once. Write your thesis in public until a partner stumbles on it. Source one deal nobody else saw. At the bottom of venture, sourcing beats analysis, and analysis is the part the model does anyway. Your edge is not a tool you can run. It is a perspective only you have, because you were somewhere specific and you paid attention.
And no, this is not all doom. IBM just tripled its entry-level hiring. CS starting salaries actually went up 7% to $81,535 for the people who landed. The outcomes are splitting hard, and LeetCode reps are not what decides which side you land on. The people getting offers can point at something they shipped. They have taste, and they can prove it.
I crossed under the old rules.
I made my own crossing on the old ladder. SWE to VC, a couple years ago. I had the junior engineering years that do not really exist anymore, the boilerplate grind that quietly taught me what good looks like, the exact grind a model would happily do for me today. So when I tell you to enter at a higher altitude, I am telling you to skip the stairs I climbed.
Frankly, I am not sure that is even possible. I am not sure I would get in today under the rules I am handing you. But I have never met anyone good who got there by typing fast. They got there by having something to say, and somewhere they had actually been.
So go be somewhere. Build the small thing. Hold the opinion. The door you were knocking on is bricked up. The wall right next to it is thinner than it looks.
— SWEdonym