You Can Do Science Without a University. You Can't Do It Alone.

I went looking for how to become an independent researcher in STEM. Not the humanities version. The science-and-code version. I found a stack of YouTube videos, pulled the transcripts, and read all ten. Some are genuine. Some are college-application hustle wearing a lab coat. Some are supervised student research calling itself independent. I kept the advice that holds up and combined it into one playbook. Then I added the part most of them skip. The myths to kill, and who gains if you believe them.

I build software with language models, and I do independent technical research on my own time. So this is not abstract for me. The computational lane these videos describe is the one I live in. Numbers in brackets point to the sources at the end. The videos are my starting point, not my only source. The debunk claims come from primary sources I name at the end, and I checked them.

Independent research, as I mean it here, is science you do outside a university or company lab, on your own time, without a professor assigning it. A student working under a professor for course credit is doing something else. That is apprenticeship. It is good. It is just not independent in this sense.

First, two different things called “independent”

A lot of what YouTube calls independent research is not. One creator describes emailing six or seven professors until one let him work on her side project for three semesters of course credit [3]. That is a student joining a professor’s work. Another shows how to land a spot in a lab at a university you do not even attend [5]. That is joining a lab. Both are fine paths. Both are apprenticeship. Neither is research outside an institution.

The real independent-of-institution example is quieter. A Cambridge maths graduate works a non-physics day job and does computational physics in his spare time, on a budget of zero [1]. A Cambridge professor answers the question “can a kid in a bedroom publish a paper?” [2]. Those two are the spine of this post. The rest add good craft advice, but half of the ten videos are really about getting into a lab or getting into college. Keep that in mind as we go. It matters for the debunk at the end.

The playbook that holds up

Pick a small question you can finish

The physics grad chose computational physics for a plain reason. The problems were isolated, tangible, and doable [1]. He was not trying to solve quantum gravity. He picked cases nobody had explored yet, where one person could actually finish something. The research coach Andy Stapleton gives the same advice as a rule. Break a big question into smaller ones until you can answer one in a week or two [8]. His test is your own anxiety. If a question makes you feel anxious because you do not know where to start, it is too big. Break it down until you can see the first step.

The two high-school creators say the same thing in their language. Narrow your scope hard [4][6]. “Biology” is too broad to be a project. Pratik’s example is leukemia. It is hard to diagnose until it has spread through the body. So a real question is whether you can diagnose it earlier, or cut the false-positive rate of a test [4]. That is a project.

Read the field, then improve something that exists

You cannot contribute to a field you do not know. The Cambridge professor is blunt about this. To make a meaningful contribution you must study the field until you know the significant results, what has been tried, what is still unsolved, and why it is unsolved [2]. That is the bar. Not a clever idea. Knowing the field.

How do you get there? The high-school creators describe a funnel. Start broad. Watch good videos, read articles, take notes in a journal [6]. Then narrow. Follow the things you keep coming back to. Read review articles, then specific papers [6]. The first paper you read will make no sense. That is normal. One creator compares it to the gym. The first set feels impossible. The gains come with reps [4]. Look up every term you do not know. Eventually you can read a paper front to back, and it gets fun.

Then start by making something a little better. This is the part that kills the genius myth before we even get to it. Science is not one Einstein having a flash. It is many labs each pushing a thing forward one to three percent, and those gains compound [4]. Your first project can be to optimize something that already exists. Make an encryption method five percent better. Make a smoke alarm two percent more effective at half the cost [4]. That is a real contribution. One creator’s idea did not arrive out of thin air. It came from reading thirty papers in a narrow field and looking up every word he did not understand [6].

Code is the open door

Here is the practical heart of it, and the part that fits this blog. You do not need a lab to do computational research. The physics grad works on materials simulations with free software and no budget [1]. Pratik says coding projects are the most accessible independent research. He says it across both his videos. You control everything, and you do not need a lab or mice [4][7]. The physics PhD student makes the same point for remote work. If a professor is a hundred miles away, pick a theoretical or computational project, not one that needs you in a lab ten hours a day [5].

Computational research is science done by writing and running code instead of mixing chemicals or building hardware. Simulations, data analysis, and modeling. The equipment is a computer you already own.

This is why independent STEM research is more possible now than the lone-genius story admits. The barrier for a whole class of problems is a laptop and the ability to code. If you cannot code yet, learn. World-class courses are free online. One creator took a full Stanford course on teaching computers to understand language, free on YouTube [6]. And if coding is not for you, Pratik points to database studies: finding something new in an existing data set, which does not always need coding [7]. That is the cheapest lab access there is.

Get the papers legally

You need to read the literature, and a lot of it is paywalled. One creator recommends Sci-Hub [6]. I am not passing that along. Sci-Hub is a pirate site. You do not need it. The legal routes are enough.

A preprint is a research paper posted publicly before, or without, journal peer review. It is free to read. It is not yet certified by anyone.

arXiv hosts preprints in physics, maths, computer science, and more, free [11]. PubMed Central and bioRxiv cover biology and medicine. The Directory of Open Access Journals (DOAJ) lists open-access journals. The browser tool Unpaywall finds legal free copies of paywalled papers. Your public library often has journal access. And you can email the author. Many authors will send you their paper. Use those. Skip the pirate site.

Run the loop, and write before you are ready

Stapleton’s core cycle is read, do, analyze [8]. Read as much as you can. Then do the experiment or run the code, even though you might be wrong and that is scary. Then analyze the result every way you can. Graphs, tables, equations. Feed what you learn back into your reading. The more cycles you fit in, the better your intuition gets.

He also says to protect the time. Block out your calendar for the core work, and treat everything else as a distraction [8]. When you work alone, nobody guards that time for you. Saying no to other people’s requests is a real option [8].

Two more of his tips belong here. Write before you think you are ready [8]. Writing shows you the gaps in your own work. If you wait until you have the whole story, you waste months. Start writing up a single graph. And double down on what works [8]. Most of your results come from a small slice of your effort. When something shows promise, follow it. When something is a dead end, stop forcing it.

Do not work alone

This is the thread that runs through every honest source. Talk about your research to anyone, until you can explain it to a child or a professor [8]. Talking shows you what you do not understand. The physics grad chats with other researchers online [1]. The biochemist says research is collaborative, not isolated. Lab mates, conferences, troubleshooting together. Papers almost always have more than one author [9].

And get a mentor or a co-author. The Cambridge professor calls a PhD an apprenticeship with a competent scientist who teaches you the trade [2]. You do not need the degree to get the apprenticeship. The high-school creator gives a concrete move. Summarize your research in a paragraph and email professors in that field, at any college, and ask to talk [7]. He says this works better than cold-asking for an internship, because you are already doing the work. The PhD student says to ask two questions before you join anything. Will I have a mentor? Is there a path to a publication? If the answer to the second is no, that is a red flag [5].

Publish honestly

Can you publish without a degree? Yes. The Cambridge professor says nothing stops a kid from submitting, and selection is on the quality of the work, not the age or affiliation of the author [2]. arXiv is open to anyone, with no fees [11].

But two things are true at once. First, the idea is the small part. The professor lists what you actually have to answer. What problem does this solve? Who benefits? Why has nobody done it? Why can you? Is the solution obvious to someone skilled in the field? Then you have to tell a good story and back it with convincing evidence [2]. The idea alone publishes nothing.

Second, there are still gates. arXiv requires an endorsement before your first submission, from an established author who confirms your work belongs in the field [11]. And arXiv is not peer review. In its own words, material is “not peer-reviewed by arXiv” and is presented as is [11].

Peer review is when other experts in the field read your paper and judge whether it is sound before a journal publishes it. It is a quality check. It is not a guarantee, and it can be faked. We will get to that.

The high-school creator adds the practical truths. A review article that only compiles existing work is not the same as an empirical paper that reports a new result [7]. Posting a paper on your personal website is not a publication. Nothing got published [7]. Getting accepted is not getting published either. Revisions take weeks to months. His first paper took about a year from acceptance to publication [7]. None of this is a weekend.

Expect failure, and keep going

The biochemist’s whole video is a reality check. Experiments do not work the first time, or the second, or the third [9]. You troubleshoot. You repeat. You wait for samples and machines. Most days are small steps, not breakthroughs. And you do not have to be a genius. Nobody knows exactly what they are doing. Everyone struggles and Googles and asks colleagues. Research is figuring things out [9].

The PhD student titled his video “Getting Into Research is Easier Than You Think.” Then he admits the truth. Research is full of dead ends and failed experiments. The hardest part is not the technical work. It is persistence [5]. Believe the ending, not the title. When something fails, you are not back at the start. You are starting from experience, because you now know one more thing that does not work [9].

The myths to kill

Now the part most of these videos skip. Some of what they sell you is wrong, and some of it is wrong on purpose, because someone makes money when you believe it.

The lone genius

The PhD coach James Hayton calls the independent researcher a dangerous myth when it means doing everything alone [10]. His point is that there is no such thing as a truly independent researcher in academia. Even Einstein, working in a patent office, was in constant correspondence with other scientists. His work became known only through that engagement [10]. Researchers are interdependent. They need other people for the work and for their own wellbeing. Isolation makes both worse [10].

The biochemist backs this with a practical pattern. Experimental papers almost always have more than one author [9]. Even the good high-school journal is built around mentoring, not solo genius [12]. So when a video sells you the picture of one brilliant person alone in a basement changing the world, it is selling you a movie. The real thing is a network. You can be outside a university. You cannot be outside the conversation.

The credential myth, and its mirror

The credential myth says you need a degree to be taken seriously. The Cambridge professor debunks that. Quality is judged, not credentials [2]. Good.

But there is a mirror myth, and it is the one the hustle videos push. It says the degree is a scam, you can skip it, go solo, and that is the shortcut. The professor flips it. Doing it solo is the opposite of a shortcut [2]. The apprenticeship, the PhD, is the shortcut, because someone teaches you the trade instead of you reinventing every mistake. And if you just want to publish one idea and then get out, do not get a PhD. Co-author with an experienced scientist who will help you shape and present it [2]. The honest version is not “credentials are everything” and not “credentials are nothing.” It is “the training matters, and you can get the training without the degree, but you cannot skip the training.”

The million-dollar lab

You do not need a million-dollar lab for computational work [1][4]. Do not let anyone tell you that you must buy your way into a lab to do science. For wet-lab work, you do need access, and the honest move is to borrow it. Use cheap materials. One creator built a solar-panel project for about twenty-five dollars [4]. Borrow from your school. Email local labs and professors. It is easier to get lab access for your own idea than to get a generic internship, because they do not have to teach you their whole research program [4]. Access is something you arrange, not something you purchase.

The college-app hustle. Who gains?

Here is where the money is. Several of the student videos are framed as getting into a top college or winning ISEF, the big international science fair [4][6][7]. That framing is not neutral. It is a sales funnel.

In May 2023, ProPublica and The Chronicle of Higher Education documented the industry behind it. At least twenty online research programs for high schoolers had sprung up, along with a cluster of journals that publish the work [13]. Their words: “A new industry is extracting fees from well-heeled families to enable their teenage children to conduct and publish research that colleges may regard as a credential” [13]. One program, Scholar Launch, starts at $3,500, and says publication “is the objective” [13].

The peer review can be a label, not a process. ProPublica contacted eight reviewers for the Journal of Student Research, which describes itself as peer-reviewed. Four said the journal had never asked them to evaluate a manuscript [13]. To be fair, ProPublica notes some had agreed but forgotten, and some work in fields the journal does not cover. The journal says it uses more than 90 reviewers and reviews take 12 to 24 weeks [13]. Still, four of eight is a striking number for a journal whose product is peer review. One of the high-school creators names that same journal as a lower-quality option [7]. The good journal he names, the Journal of Emerging Investigators, is a real nonprofit, peer-reviewed, run by graduate students [7][12]. My read: the pay-to-publish outlets sit in the tier below it.

So ask who gains if you believe that a published paper is a thing you can buy. The mentoring company gains. The journals they launch or partner with gain. The college counselor who packages the whole thing gains. You get a line on an application that an interviewer can puncture in one question. The same creator who sells the dream also tells the truth about it. He says the students with ten papers are often slicing one narrow topic into ten similar papers [7]. A guaranteed publication for a couple grand is a product. It is not science.

This does not mean all paid mentorship is a scam. A paid teacher can be legitimate. The line is whether the research is real and the publication decision is independent, or whether the publication is the thing being sold [13].

The AI shortcut

This is the one I care about most, because I build with these models every day. The promise is that AI will do your research for you. Read the papers, write the review, generate the citations. The record says otherwise.

Walters and Wilder had ChatGPT write short literature reviews. Then they checked every citation. Of the works GPT-3.5 cited, 55 percent were fabricated. They did not exist. For GPT-4 it was 18 percent [14]. Even the citations that pointed to real works often had a substantive error. Wrong year, wrong authors, wrong journal. That was 43 percent of GPT-3.5’s real-work citations and 24 percent of GPT-4’s [14]. A second study looked across disciplines. GPT-3.5’s citations referred to real works about 74.5 percent of the time. They were fully accurate far less often. Accuracy was 32.7 percent in the natural sciences and 8.5 percent in the humanities [15].

A fabricated citation is a reference an AI invents. It looks like a real paper. It has authors, a title, a journal, sometimes a DOI, the ID number a real paper carries. It does not exist.

AI is a useful assistant. It can explain a term, suggest an angle, help you code. It is not an author, and it is not a source. Every citation it hands you is a claim until you check it against the primary record. The whole playbook above depends on knowing the field. A model that makes up a fifth of your references is not helping you know the field. It is helping you fake it. And the fake is exactly what an interviewer, or a reviewer, will catch.

The honest version

You can do real STEM research outside a university. The computational lane makes it more possible than ever, on a laptop, for free. But you cannot do it alone. The idea is the small part. Knowing the field, doing the work, telling the story with evidence, and getting other people to check it. That is the large part.

Much of the “independent researcher” content is either college-application hustle or supervised apprenticeship wearing a costume. The genuine advice underneath is simple and unglamorous. Pick something small enough to finish. Read until you know the field. Start by making something a little better. Code if you can. Get the papers legally. Run the loop. Write early. Talk to people. Get a mentor. Publish honestly. Expect to fail a lot, and keep going. Check everything, especially what an AI hands you.

That is not a shortcut. It is the work. The people selling you a shortcut are usually selling you something else.

References

  1. CasualPhysicsEnjoyer, How I do independent physics research!, YouTube, October 9, 2023. Source for the zero-budget independent computational physics path, choosing tractable problems, free software, and chatting with other researchers.
  2. Frank Stajano Explains, Can I publish a research result without a degree?, YouTube, March 31, 2023. Source for quality over credentials, the idea as the small part, the PhD as apprenticeship, and solo as the opposite of a shortcut.
  3. codergio (Kota gimel), Engineering Independent Research - What is it and Why do it?!, YouTube. Source for the supervised, for-credit undergrad version of “independent research.”
  4. Pratik Vangal, How I Began Independent Research (5-Step Process), YouTube, August 15, 2023. Source for narrowing scope, the gym analogy for reading papers, incremental science, coding as the accessible lane, and cheap-materials lab access.
  5. Dario Tringali, Getting Into Research is Easier Than You Think, YouTube. Source for joining a lab, computational projects for remote work, the mentor and publication questions, and persistence as the hardest part.
  6. Rishab Jain STEM, How to Pick a Science Research Topic & Idea: FULL GUIDE, YouTube. Source for the broad-to-narrow literature funnel, note-taking, and the idea coming from thirty papers. Also the Sci-Hub recommendation this post declines to pass along.
  7. Pratik Vangal, How I Published 3 Research Papers in High School, YouTube. Source for review versus empirical papers, the mentor requirement, journal tiers, website-posting is not publication, and the ten-similar-papers trick.
  8. Andy Stapleton, How to get better at doing research [7 crazy simple tips], YouTube. Source for the read-do-analyze cycle, breaking questions down, the pub test, writing early, protecting time, and doubling down on what works.
  9. Elizabeth, What No One Tells You about Research… biochemistry, STEM, YouTube. Source for failure and waiting as the daily reality, collaboration, and not needing to be a genius.
  10. James Hayton PhD, The dangerous myth of the ‘independent researcher’, YouTube, October 3, 2022. Source for interdependence, the Einstein correspondence point, and isolation as harmful.
  11. arXiv, About arXiv and Endorsement. Source for open to anyone, no fees, moderation that is not peer review, and the first-submission endorsement requirement.
  12. Journal of Emerging Investigators, described in Incorporating Primary Scientific Literature in Middle and High School Education, Journal of Microbiology & Biology Education, 2016. Source for JEI as a nonprofit, peer-reviewed, mentoring-based youth journal.
  13. Daniel Golden and Kunal Purohit, The Newest College Admissions Ploy: Paying to Make Your Teen a “Peer-Reviewed” Author, ProPublica, co-published with The Chronicle of Higher Education, May 18, 2023. Source for the paid research-program industry, the twenty programs, Scholar Launch’s $3,500 price, and the four-of-eight Journal of Student Research reviewers.
  14. William H. Walters and Esther Isabelle Wilder, Fabrication and errors in the bibliographic citations generated by ChatGPT, Scientific Reports 13:14045, 2023. DOI 10.1038/s41598-023-41032-5. Source for the 55 percent and 18 percent fabrication rates and the 43 percent and 24 percent error rates.
  15. J. Mugaanyi, L. Cai, S. Cheng, C. Lu, and J. Huang, Evaluation of Large Language Model Performance and Reliability for Citations and References in Scholarly Writing: Cross-Disciplinary Study, Journal of Medical Internet Research 26:e52935, 2024. DOI 10.2196/52935. Source for the 74.5 percent real-citation rate and the 32.7 versus 8.5 percent accuracy split.
Written on November 20, 2026