A guide for anyone wondering what to do next · August 2026

Where the work went

Almost nobody's job is being deleted. It is being moved, renamed, split in two, or quietly stopped at the entrance. This is a map of where it moved to, what the new version is called, what it pays, and how hard it is to get.

Sajith Amma15 August 202622 min read

Every job, one list

Every job, and how to get it

Open any row for the skills, what to learn, what it pays and a link to a real advert.

Pay in solid text is what the employer published. Pay in grey underlined text is an estimate. Where a job title will not find you anything, the row tells you what to type instead.

Where the work went

Your job did not disappear. It moved.

Almost every job in that list is the next version of one that already existed. Here is the shape of the move, and the four things worth understanding before you make yours.

Almost every job here is the next version of one that already existed. Not a replacement, a continuation. The recruiter becomes the person who can tell a real ML engineer from a fluent one. The sales engineer becomes the person who demos a system that is sometimes wrong. The pattern repeats across every function.

What is actually going on

Six things worth knowing before you plan your next move. Two of them are uncomfortable and they are here anyway, because you cannot plan around something nobody tells you.

The best openings are in the least glamorous places. Some jobs are hiring hard while almost nobody is qualified for them. Licensed electricians. Liquid cooling technicians. Commissioning agents. Grid interconnection specialists. AI auditors. Several want a trade or a certificate rather than a degree, and they are wide open for a dull reason: the training takes four years and the demand arrived last year.
The title you search for is often not the title employers post. Search "multi-agent orchestration engineer" and you will find nothing, anywhere. The work is real and being hired for hard, under a different name. If your searches keep coming back empty, this is usually why, and the fix takes one minute.
The entry level is where the damage is, and it is not a layoff story. If you are starting out, you are not imagining it. Employment for developers aged 22 to 25 has fallen about a fifth since 2024. Nobody was fired. They were simply not hired, which is why there was never a headline about it.
The AI data economy split into two markets that barely touch. Basic labelling pays eight to twelve dollars an hour and is being automated away. Writing training data in a field you are actually qualified in pays eighty to a hundred and ninety. Same industry, ten times the money, and you reach them through completely different doors.
Being the person who checks the machine is now a job. Code review, evaluation engineering, escalation handling, model audit, red teaming. When output volume rises and confidence stays high, judgement becomes the scarce input. Most of these roles pay above the thing they are checking.
The compliance rush everybody priced for August 2026 moved. The EU AI Act's high risk rules were pushed back to December 2027. The jobs are real and worth moving toward. Just know that a lot of advice written this year is urging you to hurry toward a date that already moved.

One

You are searching for the wrong word

If your job search keeps coming back empty, this is the first thing to check. It is also the quickest thing on this page to fix.

"Prompt engineer" is no longer a job you can apply for. Companies stopped posting it. What they did not stop is wanting the skill, which now sits inside almost every AI job description as one line among many. If that is your strength, you are not out of luck. You are looking under a name nobody files under any more. Apply for AI Engineer, or AI Evaluation Engineer, and put the prompting on your CV as a skill.

This is the webmaster story again. The job did not vanish, it dissolved into everybody else's. And it is not the only one: below are the titles that will waste your afternoon, next to the words that will actually return results.

What to type instead. On the left, what people search for. On the right, what the job is actually filed under, which is usually a plain engineering title with the AI part left unsaid, or the name of the team rather than the role.

Searching the fashionable name returns nothing, while the work is being hired for aggressively under another name.

The most common reason a good search finds nothing

Two

If you are starting out, it is genuinely harder

This is the part nobody wants to say plainly, so here it is. You are not failing at something everyone else is finding easy. The entrance got narrower, and you can see it in the numbers.

The top bar and the two below it are measuring different things. The first counts how many young developers are in work. The others count how many graduates get hired each year, which always swings harder. Look at the last bar before you despair: engineering has held up far better than the rest of tech.

Notice what did not happen. There were no mass layoffs of junior staff. Companies simply stopped opening the door. That is why you never saw a headline about it, and it is why applying feels like shouting into a room with nobody in it. Nothing about your CV caused this.

Here is the part worth holding onto. Engineering is down about 11 percent while tech overall is down 25. Engineers are now a bigger share of everyone getting hired than they were in 2019, not a smaller one. The profession is not shrinking. The entrance to it is crowded. Those are very different problems, and the second one you can route around.

There is also a clock on this. Companies are hiring seniors to supervise AI while refusing to hire the juniors who would have become those seniors. That maths runs out somewhere around 2030, and plenty of people running these companies know it. Which means the squeeze you are living through is a phase, not the permanent shape of the industry.

Four ways in that are still open

Go where the job needs a person in the room. Forward deployed and solutions roles hire for customer sense over code volume, and some of them still run graduate intakes. Go where the skill is too new for anyone to have ten years of it. Evaluation engineering barely existed three years ago, so nobody can outrank you on experience. Go where there is an apprenticeship. The data centre build runs proper ones, with published entry criteria and no degree required. Or go around the credential entirely by building something in public, which a surprising number of employers now name as an acceptable substitute.

Three

Two data jobs, ten times apart

More new AI jobs sit here than anywhere else, and it is the corner people get burned by, because the growth everyone quotes and the pay everyone quotes belong to two different jobs.

Both of those are real, advertised, today. The bottom is falling because the machine now does the first pass and only hands you the hard cases. The top is rising because a model cannot check a diagnosis, a contract or a proof without someone who could have written it. A physics PhD is openly listed at eighty dollars an hour. A frontend engineer at ninety.

The part that matters if you are considering this: you cannot climb from the bottom to the top. They are not a ladder. The well paid tier hires you for a qualification you already have, in medicine, law, accountancy, mathematics or engineering. The other tier hires you for being available. If you hold a real credential in anything, you are being underpaid by roughly ten times unless you apply to the right one.

And this is not a niche. One company alone pays around sixty million dollars a month to more than thirty thousand people, and expects that spend to multiply several times over in the next few years. This work is quiet, it is remote, and most people qualified for the good tier of it have never heard it exists.

Four

Where the odds are actually good

Everyone tells you to follow the demand. That is half the advice, and following only half of it is how people end up applying for the same famous jobs as thousands of others. What you actually want is the job nobody has queued for yet.

Every job on this page, placed by how much hiring there is and how many people you are up against. Tap any dot to see what it is. Aim for the green box, top left: plenty of hiring, few rivals. The crowded corner on the right is where everyone is applying, and it is full of the titles you have already heard of.

The green box is not luck. Almost everything in it is scarce for one of two reasons, and both of them buy you years. Either the training genuinely takes a long time, like the electrician whose apprenticeship runs four to five years, so today's intake does not qualify until about 2030. Or the work is so new that the manuals were only written months ago, in which case nobody on earth has three years of experience in it and the person who starts this week is already near the front of the queue.

The crowded corner has a pattern too. Those are the jobs that are easy to retrain into and famous enough that everybody is retraining into them at the same time. They are still good jobs. Just know that one more course certificate will not separate you there. Only something you have actually built and can show somebody will.

Now find your own

Type what you do today and see what it turns into, with what changed and what to learn to get there. Most people find themselves on the first try.

Try:

Nothing matching? Go back to the full list, which searches skills, tools and backgrounds too.

Before you act on any of this

Every job here was read off a real advert or a named report in August 2026, and each one links back to its source so you can check it yourself before you rearrange your life around it. Where a company printed its own salary band, that is the number you see.

Not every pay figure is equally solid, so the page shows you which is which. Solid text means an employer published that band. Grey underlined text is an estimate, and the healthcare, finance, manufacturing and public sector roles are mostly estimates. Use those to judge whether a field is worth exploring, never to negotiate with.

And treat all of it as a photograph rather than a prediction. Hiring moves faster than any page can, and a salary that was live in August 2026 may not last the year. Nobody here knows what your market looks like in your city. What this can do is tell you which doors are open, what they are called, and what to learn before you knock.