by admin | Nov 13, 2026 | Opportunities
Ever wished someone would pay you to dig into one of the biggest problems in AI, even if you don’t have a PhD or a research background? That is basically what Anthropic is offering through its Fellows Program, and most people have never heard of it.
It is a paid, four month research program where you work on real AI safety or AI security problems with mentorship from Anthropic’s own researchers. No published papers required. No PhD required. Just a technical background and a willingness to learn fast. It sits alongside other genuine funded paths we have covered, like finding genuine scholarships and learning opportunities and the fully funded AI safety scholarship in Germany we wrote about recently.
What Is the Anthropic Fellows Program?
The Fellows Program is Anthropic’s way of bringing outside talent into AI safety work without forcing people through the usual academic pipeline. Fellows spend four months on an empirical research project, usually one aimed at a public output like a paper or an open source tool.
Anthropic mentors pitch project ideas first, then fellows pick one and shape it with that mentor’s guidance. The research areas cover scalable oversight, adversarial robustness and AI control, mechanistic interpretability, AI security, model welfare, and a few other tracks including economics and policy. The security track connects closely to the defensive work we covered in Anthropic’s cyber mission.
Past fellows have built agents that find smart contract vulnerabilities, studied how AI models respond to jailbreak attempts, open sourced interpretability research, and investigated how models can pick up strange behaviors from their own training data. It is hands on, not theoretical.
What You Actually Get Paid
This is the part that surprises people. Fellows receive a weekly stipend of roughly 3,850 USD (or the GBP or CAD equivalent) for about 40 hours a week, plus around 15,000 USD a month in compute funding and extra research expenses on top. That is for four months of work, with a possible extension if the project is going well.
You also get direct mentorship from Anthropic researchers, some of whom work on alignment and interpretability full time. For a lot of fellows, that mentorship ends up mattering as much as the money.
Who Can Actually Apply
Here is the part that makes this different from most research fellowships: Anthropic says it values coding ability, comfort working with AI tools, and the motivation to reduce serious AI risks more than it values credentials. No PhD, prior machine learning experience, or publication history is required.
Past fellows have come from physics, math, computer science, and cybersecurity backgrounds, not just machine learning. You do need to be able to work full time for four months and hold work authorization in the US, UK, or Canada, since Anthropic does not sponsor visas for this program.
From my own experience working across websites, cybersecurity, and online tools, this is the kind of program that rewards people who have already been building and shipping things on their own, even without a formal research title. If you have a portfolio of real projects, that often speaks louder than a transcript. If you do not have one yet, our guide on how to build an AI portfolio is a reasonable place to start before you apply.
The Numbers Behind the Program
Anthropic’s own data on this program, published on its Alignment Science blog, is worth knowing before you apply. More than 200 fellows have gone through it since early 2025. Over 65 percent of past fellows have published work that came directly out of their fellowship. More than 35 percent later joined Anthropic full time on AI safety teams, and many others moved into safety roles at other companies.
In other words, this is not just a stipend to sit on a resume. For a real share of fellows, it has turned into an actual career path into AI safety.
How and When to Apply
Applications run on a rolling basis rather than one fixed yearly deadline, and Anthropic runs several cohorts a year. For context, the window for the January 2026 cohort closed on October 18, which has already passed by the time you are reading this, but that is normal for a rolling program. A new window opens for each upcoming cohort.
That means the best move is to check the current listing directly rather than assume you missed your shot. Apply through Anthropic’s own job posting, which routes through its recruiting partner for this program. Pick one track (safety, security, or another listed workstream) rather than applying to several at once.
Important tip: watch out for scam recruiters. Anthropic’s official guidance is clear that its recruiters only contact candidates from @anthropic.com addresses, and the program never asks for payment or banking details before your first day.
Common Questions
Do I need a machine learning background to apply?
No. Anthropic says strong coding skills and the ability to learn fast matter more than prior ML experience. Fellows have come from physics, math, and cybersecurity backgrounds.
Is this program remote?
You must hold work authorization in the US, UK, or Canada and be located there during the program, since Anthropic does not sponsor visas for fellows.
What happens after the four months?
Some fellows extend their project, some join Anthropic full time, and others carry their research and skills into safety roles elsewhere. There is no guarantee either way.
Is the stipend the only funding?
No. On top of the weekly stipend, fellows get roughly 15,000 USD a month in compute funding plus research expenses, which matters a lot for anything involving model experiments.
Final Takeaway
The Anthropic Fellows Program is one of the more practical ways to get paid, real experience in AI safety without first spending years on a PhD. If you have solid technical skills and genuine curiosity about keeping AI systems safe, it is worth checking the current application window and seeing where you would fit. Even if you are not ready this round, now you know the path exists.
by admin | Nov 5, 2026 | Research & Productivity
If you’ve ever tried to use ChatGPT for a literature review, you’ve probably run into the same problem every researcher does. It sounds confident even when it’s wrong, and tracking down whether a citation is actually real can take almost as long as just reading the paper yourself.
That trust gap is exactly what Elsevier, the publisher behind journals like The Lancet and Cell, built its AI tool LeapSpace to close. It has been rolling out through 2026, and by September it had grown into something genuinely different from the general purpose chatbots most people reach for when they are trying to find or understand academic research.
What LeapSpace Actually Does
LeapSpace is not trying to be a chatbot for everything. It is built specifically for research tasks: finding relevant papers, summarizing findings, spotting contradictions between studies, and pulling together a literature review you could actually show your advisor. Ask it a question and it shows the steps it took to reach the answer, not just the answer itself, so you can check its work instead of taking it on faith.
Every claim it makes links back to a traceable source. Elsevier calls these “Trust Cards,” short explanations of why a particular source was cited and whether it contradicts something else in the results. That single feature addresses the thing that makes most researchers wary of AI tools in the first place: you usually cannot tell when they have made something up.
Built On Published Research, Not the Open Web
Where general chatbots pull from whatever is on the open web, LeapSpace is built on Elsevier’s own library plus licensing deals with other publishers. As of September 2026 it draws on more than 23 million full peer reviewed articles and books, on top of 2.6 billion cited references from Scopus covering work from over 7,000 publishers. Elsevier has also signed licensing agreements with BMJ Group, JAMA Network, Emerald Publishing, Sage, and IOP Publishing, among others, so the tool is not limited to Elsevier’s own journals.
That matters more than it sounds. Elsevier’s own research found that only 22% of researchers trust existing AI tools, mostly because general models are trained on a mix of sources with no clear way to tell what is reliable. Full text access instead of just abstracts means LeapSpace can actually read a paper’s methodology section, not just guess at it from a summary.
Who Can Use It, and What It Costs
Institutions and companies can already buy LeapSpace, and individual plans for academics and students became available earlier in 2026, including a free access tier. If you are a student working through a literature review or a researcher trying to clear a backlog of papers, it is worth checking whether your university library already has it. A lot of institutional subscriptions quietly include tools like this without anyone announcing it loudly.
Quick tip: before you adopt any new research tool, ask whether it shows its sources clearly enough that you could defend a citation in front of your advisor. If it cannot, you are still doing the fact checking work yourself, just with extra steps.
Why This Fits a Bigger Pattern
From my own experience working with online tools and research adjacent projects, the tools that stick around tend to be the ones that make their limits visible instead of hiding them. An independent five person advisory board, including researchers from RWTH Aachen and the University of Wisconsin-Madison, now reviews how LeapSpace ranks sources and flags bias, which is a step most AI tools skip entirely.
If you are already using other AI research tools, LeapSpace is worth comparing on exactly this point: not how fluent the answers sound, but how easy it is to verify the ones that matter. Our guide on how to fact-check AI research answers walks through a simple process for that, and if you are weighing LeapSpace against something built more for speed than rigor, our Elicit vs Gemini Notebook comparison covers that tradeoff. If you want the bigger picture on using AI for coursework or your own research, our guide to AI for research and productivity is a good starting point, and our piece on Google Scholar Labs covers a free alternative worth knowing about too.
Common Questions
Is LeapSpace free?
Individual academics and students get access to a free tier, though fuller paid plans unlock more. Institutions and companies pay for broader access.
Does LeapSpace replace Google Scholar?
Not really. Google Scholar is a search engine for finding papers. LeapSpace is built to read, summarize, and cross-check papers once you have found them. Many researchers will likely end up using both.
Can I actually trust what LeapSpace tells me?
More than a general purpose chatbot, since every answer links to a specific source and flags contradictions. But “more trustworthy” is not the same as “always right,” so checking the underlying paper is still worth the extra minute.
Final Takeaway
AI research tools are multiplying fast, and most of them are optimized to sound helpful rather than to be checkable. LeapSpace is a useful reminder that the second one matters more. If you do any kind of literature review work, even occasionally, it is worth five minutes to see whether your institution already has access.
by admin | Nov 4, 2026 | Future Jobs
Picture two people at the same company, with the same Copilot or ChatGPT license, doing roughly the same job. A year from now, one of them is noticeably ahead: producing work they couldn’t have touched before, trusted with bigger decisions, and somehow never behind on anything. The other one is still using AI the same way they did on day one, mostly for quick drafts and the odd summary. Microsoft just published research trying to explain that gap, and the answer has very little to do with which AI tool either of them uses.
The company’s 2026 Work Trend Index surveyed 20,000 AI-using knowledge workers across 10 countries between February and April 2026, and combined that with real usage data from Microsoft 365 Copilot. The headline finding is that a small group of workers, Microsoft calls them “Frontier Professionals”, are getting dramatically more value out of AI than everyone else, and it’s not because they got a better model.
What People Are Actually Using AI For
Before getting into who’s winning, it helps to know what people actually do with AI day to day. Almost half of Copilot conversations (49%) are about cognitive work: analysis, problem-solving, evaluating options. The rest splits between working with other people (19%), producing finished work (17%), and finding information (15%). Most people are using AI as a thinking partner first and a content generator second. That matches what I’ve seen running websites and online tools myself: the biggest time savings rarely come from asking AI to “write this for me.” They come from using it to think through a problem before you touch the keyboard.
Meet the “Frontier Professionals”
Microsoft’s survey found that 16% of AI users fall into a category it calls Frontier Professionals. These are the people who use AI agents for multi-step workflows, build or customize multi-agent systems, and actively redesign how their work gets done instead of just bolting AI onto their old process.
The numbers behind that label are striking. Across all AI users, 58% say they’re now producing work they couldn’t have a year ago. Among Frontier Professionals, that jumps to 80%. Overall, 66% of AI users say they have more time for high-value work. The gap isn’t in access to AI. Everyone in this survey already had it.
It’s Not the Tool, It’s the Habits
Here’s the part that surprised me most: 86% of all AI users treat AI output as a starting point, not a finished answer. That’s already a healthy habit. But Frontier Professionals take it further. They’re more likely to deliberately work without AI on purpose (43% versus 30% of everyone else), and far more likely to pause and consciously decide whether a task should go to AI or stay with a human (53% versus 33%).
When Microsoft asked what skills matter most for working well with AI, the top two answers weren’t technical at all: quality control of AI output (50%) and critical thinking (46%). Not prompt engineering. Not knowing the newest model. Just the judgment to check the work and know when to trust it.
Quick tip: if you want to move toward this group, the fastest habit to build isn’t a new tool, it’s a five-minute pause before you accept any AI output: ask yourself whether this was actually the right task to hand off, and read the result like you’d check a junior colleague’s work, not like you’re grading your own.
Managers Matter More Than the AI Does
A separate study of 1,800 workers in the same report found that when managers visibly model AI use themselves, their teams report a 17-point lift in the value they get from AI, a 22-point lift in critical thinking about AI output, and a 30-point lift in trust in agentic AI. Frontier Professionals are also far more likely to have managers who set clear quality standards for AI-assisted work (83% versus 57% for everyone else) and who create space to experiment (84% versus 61%).
Teams built around Frontier Professionals also share what they learn. 61% say their team actively shares AI learnings, compared to 36% elsewhere, and they’re almost twice as likely to have documented, repeatable AI workflows at the organizational level (25% versus 14%). None of that needs a bigger AI budget. It needs a manager who talks about it openly and a team that treats “how we use AI” as something worth discussing out loud, the same way working with AI agents at work is already reshaping how teams divide tasks between people and AI.
Why Most Companies Aren’t There Yet
Microsoft’s report is honest about how early this still is. Culture, manager support, and talent practices account for more than twice the impact on AI results (67%) compared to an individual’s own mindset and effort (32%). Yet only 26% of AI users say their leadership is clearly and consistently aligned on an AI strategy. Meanwhile 65% fear falling behind if they don’t adapt quickly, but only 13% say they’re actually rewarded for trying new AI approaches when results fall short. That’s a real tension: people feel pressure to experiment, but most workplaces still punish the failed experiments that experimentation requires. It lines up with what Workday’s own 2026 workforce research found, that employers increasingly expect AI to reshape jobs rather than simply cut them, which you’d expect if the real bottleneck is how organizations adapt, not whether the technology works. See our earlier piece on how AI is rewriting jobs, not just cutting them.
What This Actually Means for Your Job
You don’t need permission from your employer to start building these habits. Microsoft also found that employers have created at least 1.3 million new AI-related roles over the past two years, according to LinkedIn data cited in the report, things like AI engineers and data annotators that barely existed before. The people filling those roles, and the people becoming Frontier Professionals inside roles that already exist, are mostly doing three things: using AI for genuinely multi-step work instead of one-off prompts, checking output critically instead of accepting it, and talking openly with colleagues about what’s working. If you’re thinking about how AI skills translate into something you can actually show an employer, it’s worth reading what we found in AI skills on your CV and what the research says they’re actually worth, and for the bigger picture on where jobs are headed, see how AI is changing future jobs.
Common Questions
What is a “Frontier Professional” exactly?
It’s Microsoft’s term for the 16% of AI users in its 2026 survey who use AI for multi-step workflows, build or adjust multi-agent systems, and actively redesign their own work process around AI, rather than using it for occasional one-off tasks.
Do I need a technical background to become one?
No. The report’s top two skills for working well with AI are quality control and critical thinking, not coding or prompt engineering. Most of the gap comes from habits, not technical skill.
Is this just about Microsoft Copilot?
The survey was built around Microsoft 365 Copilot usage data, but the behaviors it describes, pausing to decide what to hand off, checking output critically, sharing what works with your team, apply to any AI tool.
Does using AI more just mean doing more work?
Not according to the report. 66% of AI users said it freed up time for higher-value work rather than simply adding more tasks, though that benefit was strongest among Frontier Professionals who also deliberately chose when not to use AI.
Final Takeaway
The gap between people who get real value from AI and people who don’t isn’t about access to better tools anymore, almost everyone in this survey already had the same ones. It’s about habits: using AI for real multi-step work, checking its output like you’d check anyone else’s, and talking about it openly with the people around you. Those are things you can start doing this week, with whatever AI tool you already have.
by admin | Nov 2, 2026 | AI Tools
If you have used Google’s NotebookLM to summarize lecture notes or dig through a stack of PDFs, you might have noticed something different lately. The tool now goes by Gemini Notebook, and over the past couple of months Google has quietly turned it into something closer to a full study assistant.
I spend a lot of my own time testing tools like this for client work and for my own online projects, and Gemini Notebook is one of the few AI tools that actually earns its place on a student’s laptop rather than just adding another tab to close. The new features make that even more true, and there is a free offer buried in the announcement that is worth knowing about before it expires.
What changed in Gemini Notebook
Google started rolling out these Gemini Notebook study tools in mid September 2026, and most of them are aimed squarely at students rather than researchers writing papers.
- Real time voice conversations. On the mobile app, you can now talk to your notebook out loud, in close to 100 languages, and interrupt it mid sentence the way you would a person. Everything it says stays grounded in the sources you uploaded, so it is not only chatting from general knowledge. This started with Google AI Ultra subscribers, with Pro and other plans following.
- A built in audio recorder. The mobile app can record a lecture or your own spoken notes, then drop that recording into your notebook alongside your other sources. You can chat with it, search it, and cite it like any other document.
- Interactive learning overviews. Under the Reports section, Gemini Notebook can now combine a written summary with an infographic, a quiz, and flashcards in one view, instead of making you jump between separate tools.
- New quiz formats. Short answer, multiple select, and fill in the blank questions are joining the existing flashcard style quizzes. You can ask the notebook chat how you did and where you are weak, and it will point you back to the exact source material.
- Short video overviews. These are roughly one minute videos that explain a formula, diagram, or concept with narration and simple animation, available in more than 80 languages. A good one can double as a study aid for your whole class group chat.
The part students should not miss
Along with the feature rollout, Google is giving US college students a full year of Google AI Pro for free, which normally lists at $19.99 a month, with four times the usual Gemini Notebook usage limits. Students in more than 140 other countries get a year of Google AI Plus free instead, with double the usual limits.
Important tip: the offer needs a valid payment method on file and switches to a paid plan automatically once the free year ends, and the redemption deadline is December 31, 2026, so if you want it, set a calendar reminder to cancel or downgrade before the first real charge hits.
This is a genuinely useful offer if you are already studying with AI tools, but it is also exactly the kind of deal worth reading the fine print on first. Our guide on free AI student offers walks through how to spot the ones that quietly turn into a subscription you forget about.
How this fits into a normal study routine
None of this replaces actually learning the material, and Gemini Notebook is upfront that its answers stay grounded in your own sources rather than invented. That is still the safest way to use any AI study tool. Feed it your lecture slides, your textbook PDF, and your own notes, and let it summarize and quiz you on what is actually there, rather than asking it general questions it might get wrong.
If you are new to the tool, our beginner’s guide to NotebookLM covers the basics, and if you are trying to decide between Gemini Notebook and a research tool like Elicit, this comparison breaks down which one fits which job. For a wider look at research tools generally, see our guide to AI research tools for students.
Common Questions
Is Gemini Notebook the same as NotebookLM?
Yes. Google renamed NotebookLM to Gemini Notebook as part of a wider push to bring its AI products under the Gemini name. The underlying tool and your existing notebooks are not affected by the rename.
Do I need to pay to use any of these new features?
Most of the new study tools, including interactive learning overviews and the new quiz formats, are rolling out to all users over several weeks at no extra cost. The real time voice conversations started with Google AI Ultra subscribers before expanding to other plans.
How do I claim the free year of Google AI Pro or Plus?
Check Google’s own Google AI plans page for your region’s current student offer and eligibility rules, since terms can change. You will typically need a valid school email or other proof of enrollment and a payment method, and the offer must be redeemed by December 31, 2026.
Will my recordings and notes be private?
Google’s own documentation says your notes and sources stay tied to your own notebook. As with any AI tool, avoid uploading material that is confidential or that you do not have the right to share, and check your school’s own AI policy if you are unsure.
Final takeaway
Gemini Notebook was already one of the more trustworthy AI study tools because it sticks to your own sources instead of guessing. The new voice chat, quizzes, and video overviews make it genuinely faster to study with, and the free year of Google AI Pro or Plus for college students is worth claiming while it lasts, as long as you remember to cancel before the renewal if you do not want to keep paying. Read Google’s own announcement for the full list of regions and terms before you sign up.
by admin | Nov 1, 2026 | Opportunities
If you have ever looked at AI PhD programs abroad and quietly closed the tab because of the cost, there is a program in Germany worth knowing about. It pays you to study AI, not the other way around, and the next round of applications is expected to open in just a few weeks.
It is called the Konrad Zuse School of Excellence in Reliable AI, or relAI for short. It is one of several “Zuse Schools” run through Germany’s DAAD (the German Academic Exchange Service), backed by federal funding, and it is built around a question that does not get nearly enough attention in AI education: not just “can we make AI powerful,” but “can we trust it.”
What Is relAI, Exactly?
relAI is a joint program between the Technical University of Munich (TUM) and LMU Munich, two of Germany’s top research universities. Instead of treating AI reliability as an afterthought, the whole program is built around it: safety, security, privacy, and making sure AI systems behave the way they are supposed to in domains where mistakes actually matter, like healthcare or robotics.
Students work alongside relAI Fellows (senior researchers in the field), get mentoring, attend dedicated training, and become part of a growing international research network rather than just sitting in lectures.
What You Actually Get
This is the part most “free scholarship” articles gloss over, so here are the real numbers, straight from relAI’s own site:
- Master’s students receive a monthly scholarship of €934, plus extra allowances if you have a partner or children, for two years.
- Doctoral researchers (PhD) are employed on a full TV-L E13 contract for three years. That is not a stipend, it is a real German public-sector salary with the benefits that come with it, one of the more competitive packages you will find for AI PhD positions anywhere.
There is also a path to become an “associated” relAI PhD student if your funding comes from elsewhere but your research fits the program, which still gives you access to the same curriculum and community.
What You’ll Actually Study
relAI organizes its research into five areas: mathematical and algorithmic foundations, learning and instruction, algorithmic decision-making, medicine and healthcare, and robotics and interacting systems. That healthcare track matters more than it might sound. A lot of the current debate around AI in medicine comes down to explainability, whether a doctor can actually understand why a model reached a conclusion before trusting it with a diagnosis. That is squarely in relAI’s territory.
From my own experience working around websites, cybersecurity, and AI tools, the part of this field that gets the least attention is also the most important one. Everyone wants to talk about how powerful a model is. Almost nobody asks how you would know if it quietly started getting things wrong. Programs like this exist because that question needs real researchers working on it, not just a disclaimer at the bottom of an app.
Quick tip: if you are even considering applying, start lining up your two recommendation letters and your CV now. Strong applications need time, and scrambling for references in the last week never goes well.
Who Can Apply
You do not need to already be based in Germany or the EU. The program accepts applicants from anywhere, as long as you meet the academic bar:
- An excellent bachelor’s degree (for the MSc track) or master’s degree (for the PhD track) in computer science, mathematics, engineering, natural sciences, or a related data science or machine learning field.
- A genuine interest in reliable AI, meaning safety, security, privacy, or responsibility in AI systems.
- English proficiency at C1 level, or proof that your previous degree was taught mainly in English.
- For the PhD track specifically, you will also need to name two preferred supervisors from relAI’s Fellows and explain why, plus two referees willing to write recommendation letters.
When to Apply
relAI typically opens its PhD call in the fall or winter and its master’s call in spring or summer. The most recent PhD round closed on January 13, 2026, and relAI’s own site says to expect the next call in December 2026, which is right around the corner. If you are aiming for the PhD track, now is genuinely the time to prepare, not wait until the call opens.
This is also just one piece of a bigger picture. Germany’s DAAD runs several of these Zuse Schools across different AI specialties, and relAI is sitting alongside other fully funded routes into AI research worth knowing about, including some of the fully funded AI master’s programs across Europe we covered recently. If you are early in your search, our guide on finding genuine scholarships and learning opportunities online is a good place to start, and if the idea of “reliable AI” is new to you, our plain-English explainer on what explainable AI actually means fills in the background.
Common Questions
Is this only for German or EU citizens? No. relAI accepts applicants internationally, as long as you meet the academic and language requirements.
Do I need research experience already? A strong academic record and a genuine interest in the field matter most. For the PhD track, having a specific supervisor and thesis idea in mind will strengthen your application.
What does “reliable AI” actually mean? It covers making AI systems safe, secure, privacy-preserving, and resistant to manipulation, essentially everything that needs to be true before AI can be trusted in high-stakes settings like medicine or infrastructure.
Is there support if I have a family? Yes, master’s scholars receive additional allowances for partners and children on top of the base scholarship.
Final Takeaway
Fully funded AI programs that actually pay a livable salary, rather than a token stipend, are rare. relAI is one of them, and it is specifically built for people who want to work on making AI trustworthy, not just bigger. If that sounds like the kind of problem you want to spend a few years on, bookmark the official application page now, start preparing your documents, and check back as December approaches.
