by admin | Oct 16, 2026 | Research & Productivity
If you have ever watched a university lab wait weeks for a bigger cloud-computing budget just to run one more experiment, you already know the real bottleneck in science is rarely a shortage of ideas. It is access. On October 15, 2026, OpenAI tried to chip away at that problem for a specific group of researchers with a new program called the Rosalind Academic Plan.
If you work in a lab, teach at a university, or just follow how AI is changing research, here is what the plan actually offers, who qualifies, and why it is worth paying attention to even if you will never apply for it yourself.
What Is the Rosalind Academic Plan?
The Rosalind Academic Plan is a new OpenAI program that gives eligible U.S. university labs free access to its commercial AI models, paired with GPT-Rosalind, OpenAI’s specialized model for life sciences, plus a workspace called Rosalind Workbench. Until now, GPT-Rosalind access had mostly gone to large companies and select government partners, so opening a path for academic labs is a genuine shift.
Each approved lab can activate up to 30 named seats at no charge, for up to 12 months, with usage limits attached. The offer covers subscription-style seats only. It does not include API access, so labs cannot wire the models directly into their own custom research software.
What Is GPT-Rosalind, and Why Isn’t It Automatic?
GPT-Rosalind launched back in April 2026 as the first model in a new OpenAI series built for scientific work. It is named after Rosalind Franklin, whose X-ray images were central to discovering the structure of DNA. OpenAI designed it to reason about molecules, proteins, genes, pathways, and disease biology, and to handle multi-step research tasks like literature review, experiment planning, and data analysis using scientific tools and databases. Early partners applying it to real workflows reportedly include Amgen, Moderna, the Allen Institute, and Thermo Fisher Scientific.
Here is the part that trips people up: getting a seat under the Rosalind Academic Plan does not automatically unlock GPT-Rosalind itself. That model needs a separate trusted-access review, on top of the lab’s approval for the plan. OpenAI is clearly treating a biology-tuned reasoning model as higher-stakes than a general chatbot, which tracks with how cautious the field has been about models that touch drug discovery and pathogen-adjacent research.
Who Can Actually Apply
Eligibility is reviewed at the lab level, not the individual level. OpenAI looks at:
- The lab lead’s identity, university appointment, and authority to oversee access
- The lab’s university affiliation and current research activities
- The intended use of the models and data
Labs at universities that already have an OpenAI Enterprise agreement apply through their institution. Labs at universities without one go through OpenAI’s own trusted-access route instead. OpenAI also states that research data shared through the plan is not used to train its models by default, which matters a lot if your work involves anything sensitive or unpublished.
Tip: before you plug any real experimental or patient-derived data into a new research AI tool, read the data-use and training policy yourself rather than trusting a headline summary of it. Policies like this get updated, and “not used to train by default” can still mean something different from “never touches our servers.”
The Bigger Pattern Behind the Headline
From my own experience working around websites, online tools, and digital projects, I have noticed that AI research access is splitting into two very different tracks right now. Broad discovery tools are opening up fast and wide, the way Google Scholar Labs expanded its AI research features to more users this year. But specialized, higher-stakes models like GPT-Rosalind are moving the opposite direction: more gated, more reviewed, more careful about who gets in.
That split makes sense once you think about what is at stake. A general research assistant that helps you summarize papers is a very different risk profile from a model fine-tuned to reason about disease pathways and drug design. If you are comparing everyday research tools rather than specialist lab software, our recent look at Elicit vs Gemini Notebook is a useful next read.
It is also a reminder that whatever AI tool you are using for research, the output still needs your judgment on top. That applies just as much to a specialized biology model as it does to a general chatbot summarizing a journal article. We covered the practical side of that in how to fact-check AI research answers before you cite them, and it is worth revisiting before any serious project, academic or otherwise.
What This Means for You
If you are not affiliated with a U.S. university lab, this specific program will not reach you directly. But it is still a useful signal: AI companies are increasingly willing to subsidize serious research use cases while keeping tighter control over who touches their most specialized models. If you are a student or early-career researcher, it is worth asking your supervisor or department whether your institution already has an OpenAI Enterprise agreement, since that changes which door you knock on.
Common Questions
Is the Rosalind Academic Plan free forever?
No. Approved labs get up to 30 seats free for up to 12 months, with usage limits. OpenAI has not said what happens after that period.
Does my university need an existing deal with OpenAI first?
Not necessarily. Universities with an OpenAI Enterprise agreement apply through their institution. Universities without one can apply through OpenAI’s separate trusted-access route.
Can an approved lab use GPT-Rosalind right away?
No. A seat under the plan covers OpenAI’s general commercial models. GPT-Rosalind itself needs an additional, separate trusted-access review.
Does the plan include API access for custom tools?
No. It covers subscription seats only, not API access, so labs cannot connect the models directly into their own research software through this plan.
Final Takeaway
The Rosalind Academic Plan is a narrow program on paper, limited to U.S. university labs, capped seats, and a one-year clock. But it says something broader about where AI in research is heading: more free access for people doing real scientific work, and more scrutiny over who gets the most powerful, specialized tools. Whether you are in a biology lab or just reading a paper with an AI assistant open in another tab, the same habit still applies: use the tool, then check its work yourself.
by admin | Oct 15, 2026 | Research & Productivity
You open twelve browser tabs, a stack of PDFs, and a half-finished spreadsheet, and you still cannot answer a simple question: what does the research actually say? If that sounds familiar, you have probably already tried an AI research tool to speed things up. Two of the most talked-about options right now are Elicit and Gemini Notebook, the tool most people still know by its old name, NotebookLM.
Both tools use AI to help you work through research faster, but they are built for different jobs. Picking the wrong one wastes time you were trying to save. Here is a plain-English look at what each one actually does, where it falls short, and which one fits your situation.
What Elicit Is Really Good At
Elicit is built for people doing serious literature work: students writing a thesis, researchers running a systematic review, or anyone who needs to find and summarize academic papers quickly. It searches across more than 138 million papers and over 500,000 clinical trials using semantic search, so you do not need to guess the exact keywords a paper might use.
- A Research Agent that gathers evidence from journals, clinical trials, and your own uploaded documents, then builds cited reports, tables, and even slides
- Systematic review support that automates screening and data extraction, with Elicit’s own figures suggesting researchers can save a large share of the time that step normally takes
- A Library to store and reuse sources across projects, plus Routines that keep you updated when new relevant research is published
- Sentence-level citations on AI-generated claims, so you can trace every summary back to its source
Elicit has a free Basic plan for casual searching and reading, with paid tiers that unlock exports, larger systematic reviews, and API access for heavier workloads. If your work involves citing sources properly, our guide on how to fact-check AI research answers before you cite them is worth reading alongside any tool you choose, Elicit included.
What Gemini Notebook (Formerly NotebookLM) Does Differently
Google renamed NotebookLM to Gemini Notebook in July 2026, though plenty of people still search for it by the old name, so we will use both here. The core idea has not changed: you upload your own sources, PDFs, notes, slides, websites, and Gemini Notebook answers questions strictly grounded in what you gave it, with citations pointing back to the exact passage.
- Source-grounded chat that will not wander off into made-up answers outside your uploaded material
- Audio and video overviews that turn your notes into a spoken discussion or a short explainer video
- Mind maps, flashcards, quizzes, and slide decks generated automatically from your sources
- A secure code execution environment per notebook, useful for data analysis grounded in the files you uploaded
- Syncing between the Gemini app and the standalone Gemini Notebook experience
Where Elicit goes out and searches the wider body of academic literature for you, Gemini Notebook stays inside the walls of whatever you feed it. That difference decides which tool actually helps you.
Quick tip: if you are not sure which job you are doing, ask yourself this: am I trying to find research I have not read yet (use Elicit), or am I trying to understand and organize research I already have (use Gemini Notebook)? That one question settles most of the confusion.
Elicit vs Gemini Notebook: Practical Scenarios
- Writing a literature review from scratch and need to discover relevant papers: Elicit
- You already have a folder of PDFs, lecture slides, or meeting notes and want a study companion: Gemini Notebook
- Running a formal systematic review that needs PRISMA-style screening: Elicit
- Turning dense research into an audio summary you can listen to on a commute: Gemini Notebook
- Extracting data into a table across dozens of papers: Elicit
- Building a quiz or mind map from your own course material to revise for exams: Gemini Notebook
From my own experience working with online tools and digital projects, the tools that save real time are the ones matched to the job in front of you, not the ones with the longest feature list. Match the tool to the task first, then worry about which one looks nicer.
Which One Should You Actually Pick?
If you are starting a research project and do not yet know what has already been published, start with Elicit. It will surface papers you would never have found by scrolling through Google Scholar manually, and the citation trail keeps your work honest. For deeper background on how these grounded-answer systems work under the hood, see our explainer on what RAG in AI actually means, since both tools lean on similar retrieval techniques.
If you already have the material, your own PDFs, lecture notes, or a pile of saved articles, and you want to understand, organize, and revise it faster, Gemini Notebook is the better fit. Our guide on how to chat with a PDF using AI covers a similar workflow if you want more free and low-cost options in the same space.
Plenty of students and researchers end up using both: Elicit to find and screen the literature, then Gemini Notebook to turn the papers they have collected into something they can actually study from. If you want a wider view of how AI fits into a research workflow beyond these two tools, our piece on how AI can help with research and productivity is a good next read.
Common Questions
Is Elicit free to use? Yes, Elicit has a free Basic plan with unlimited search across its paper database and limited use of its Research Agent and Reports features. Paid plans add exports, larger systematic reviews, and higher usage limits.
Did NotebookLM really change its name? Yes. Google announced the rename to Gemini Notebook in July 2026. It is the same underlying product, now more tightly connected to the Gemini app and Google Search.
Can I trust the citations these tools give me? Both tools link claims back to source text, which is a good starting point, but you should still open the original source before citing it in your own work. Our guide on fact-checking AI research answers walks through exactly how to do that.
Final Takeaway
Elicit and Gemini Notebook are not really competitors, they solve two different problems. Elicit helps you find research you have not seen yet. Gemini Notebook helps you make sense of research you already have. Figure out which problem you are facing today, pick the matching tool, and you will get more done in the next hour than an afternoon of random tab-switching ever managed.
by admin | Oct 14, 2026 | Future Jobs
You finish a report in half the time because an AI tool drafted most of it. That feels like a win, and most days it is. But here is a question worth sitting with: if you stopped using that tool tomorrow, could you still do the whole job yourself, at the same level, without it?
For a growing number of workers, the honest answer is “not as well as before.” That gap has a name now: AI deskilling. It is quietly becoming a bigger talking point among economists than the old fear of mass AI layoffs, and it is worth understanding properly instead of just scrolling past the headline.
What AI Deskilling Actually Means
Deskilling is not about losing your job to a robot. It is about a job slowly losing the harder parts that used to make someone genuinely skilled at it. If an AI tool now does the tricky 20% of your work, the 80% left for you can end up being the easier, more repetitive part, and the muscle you built for the hard part can fade from disuse.
Anthropic, the company behind the Claude AI models, looked at this directly in its newest Economic Index report, published in early 2026. The researchers measured millions of real Claude conversations and ranked tasks by how much education they typically require. The pattern they found was not subtle.
What the Data Actually Found
Claude sped up college-level tasks by roughly 12 times compared to doing them manually, versus about 9 times for tasks that only need a high school level of skill. In plain terms, AI is giving the biggest boost to the hardest, most specialised work, not the simple stuff.
Success rates told a related story. Claude completed about 66% of college-level tasks successfully, against 70% for easier tasks. Slightly lower, but not low enough to cancel out that huge speed advantage. So people keep leaning on it for the hard parts, again and again.
The researchers then ran a thought experiment: if you removed every task Claude currently handles well from a typical job, what would be left? On average, the remaining work skewed toward lower-education tasks. The report is careful to call this a signal, not a prediction, but it is exactly the deskilling pattern described above, shown in real usage data rather than guesswork.
- Data entry keyers and radiologists came out more affected once success rates were factored in.
- Teachers and software developers came out less affected than raw task coverage alone would suggest.
- Augmentation (AI helping with part of a task) is now more common than full automation, at 52% versus 45% of conversations.
The Job Losses Didn’t Show Up the Way Everyone Feared
Here is the part that should calm some nerves. A separate McKinsey survey published in September 2026 found that 32% of businesses expected AI to cut their headcount, but only 14% actually reported a reduction a year later. Two thirds of companies said total employment barely moved.
So the mass layoff wave many people braced for mostly has not arrived. That is genuinely good news. But it also means the more interesting story is happening quietly inside jobs that still exist, in the shape of what those jobs now ask of the people doing them. This is a case where two reports from two different organisations line up: fewer people are being pushed out, while the actual content of the work is shifting underneath them.
Important tip: before letting an AI tool handle a task you are still learning, try doing it yourself first at least once a week. You keep the skill, and you also get much better at spotting when the AI’s answer is wrong.
How to Protect Your Own Skills
From my own experience managing websites, online tools and cybersecurity work, the tools that save the most time are also the easiest ones to lean on too hard. I have caught myself letting an AI assistant handle a config or a script I used to write from scratch, and had to deliberately slow down and work through it manually again so the underlying skill did not quietly disappear.
A few habits actually help here:
- Use AI to check your work after you attempt it yourself, not only to produce the work from scratch.
- Keep doing the hardest 10% of a task by hand on purpose, even when a tool could do it faster.
- Ask the AI to explain its reasoning, not just give you the answer, so you are learning alongside it.
- Rotate between tasks you let AI handle and tasks you deliberately do unassisted, so no single skill goes untouched for months.
This matters most for people early in their careers. If you are a new graduate already facing a tougher entry-level job market, skipping the slow, difficult version of a task because AI can do it faster means you never actually build the judgment that senior roles depend on later. The fast way out can quietly become the slow way up.
If you work alongside AI tools day to day, it also helps to understand how AI agents actually operate at work, so you know exactly which parts of a task you are handing off versus which parts you are still responsible for. And if you are updating your CV or thinking about how to present your skills, it is worth reading about how to list AI skills on your CV in a way that shows real judgment, not just tool familiarity.
Common Questions
Is AI deskilling the same as AI taking your job?
No. Deskilling is about a job changing shape while you keep it, not about losing it. The McKinsey data above shows actual job losses have been smaller than feared so far.
Which jobs are most at risk of deskilling?
Anthropic’s data points to roles with a lot of repeatable, well-documented tasks, like data entry and certain parts of radiology, as more exposed once you factor in how reliably AI completes that work.
Can deskilling be reversed once it happens?
Largely yes, with deliberate practice. Skills fade from disuse, not permanently. Doing the harder version of a task occasionally, even when a tool could do it faster, keeps the underlying ability alive.
Does this apply to students too?
Yes, arguably even more. If you lean on AI for every assignment’s hardest part, you reach the job market with a tool-shaped gap in the exact skills employers are testing for in interviews.
Final Takeaway
AI is not quietly emptying out the job market the way a lot of headlines predicted. What it is doing, according to the people who built the models, is reshaping what is left of each job once the hardest parts get handed off. That is a manageable problem if you notice it early. Keep practising the difficult 10%, ask your tools to teach you rather than just finish for you, and you get the speed without losing the skill underneath it. For more on how AI is reshaping the workplace, see our full guide on how AI is changing future jobs.
by admin | Oct 13, 2026 | AI Guides
If you have ever used an AI chatbot and felt it answer almost instantly, there is a good chance a smaller, faster model was doing the work behind the scenes. On October 7, 2026, Anthropic released Claude Haiku 5.5, and it is built almost entirely around that idea: speed and cost, without giving up too much of what makes Claude useful.
This guide breaks down what Claude Haiku 5.5 actually is, what changed from the last version, and why it matters even if you never write a line of code. If you are still getting comfortable with AI basics, our simple explanation of what AI is is a good place to start first.
What Is Claude Haiku 5.5?
Claude Haiku 5.5 is the newest and smallest model in Anthropic’s Claude family, sitting below Claude Sonnet 5.5 and Claude Opus 5.5. Anthropic describes it as the cheapest, fastest, and most capable small model the company has released so far, according to its official announcement.
Small does not mean weak here. Haiku models are built for high volume work such as sorting support tickets, pulling data out of documents, routing requests to the right tool, or running as a background “subagent” inside a bigger AI system. Haiku 5.5 is meant to handle that kind of repetitive, large scale work cheaply and quickly, while the bigger Sonnet and Opus models take on the harder reasoning tasks.
What Changed From Haiku 4.5
A few numbers tell most of the story. Haiku 5.5 now supports a 1 million token context window, up from 200,000 tokens in Haiku 4.5. In plain terms, that means it can read and “remember” a much longer document, codebase, or conversation in one go, roughly the length of several long novels at once.
Maximum output also doubled, from 64,000 to 128,000 tokens, so it can generate longer responses in a single pass. It is also the first Haiku model with adjustable thinking effort, meaning developers can tell it to think longer on harder questions and move fast on simple ones, with medium effort set as the default.
Pricing is where it gets interesting. For prompts up to 100,000 tokens, Haiku 5.5 costs $0.10 per million input tokens and $0.50 per million output tokens, down from $1 and $5 for Haiku 4.5. Anthropic says it costs about 75 percent less to run on average than the previous version, details confirmed on its model pricing and overview page. One honest caveat worth knowing: the new tokenizer counts text slightly differently and produces roughly 30 percent more tokens for the same content, so the real world savings land closer to 85 to 90 percent rather than a flat 90 percent. Still a big drop either way.
Why This Matters Even If You Are Not a Developer
You will probably never type a prompt directly into “Haiku 5.5.” What you will notice is that AI features built on top of it, inside customer support chat, email tools, research assistants, and browser extensions, can become faster and cheaper to run. When a model gets this much more affordable, companies tend to add more AI features rather than fewer, because the cost barrier just dropped. We saw a similar shift when Claude for small business got its last major upgrade, where lower friction led to more teams actually trying it.
From my own experience working with websites and online tools, this is usually how real change reaches everyday users. It rarely shows up as “a new AI model.” It shows up weeks later as “why did this app suddenly get a smarter search bar.” The model underneath quietly changed first.
Quick tip: if you build or manage anything with AI chat features, like a website FAQ bot, a document summarizer, or an internal support tool, ask your developer or vendor whether they plan to move to Haiku 5.5. It is not a drop in replacement for Haiku 4.5, since some older API settings now return errors, so it needs a short migration, but the cost savings can be worth it fast.
How It Fits Into the Bigger Claude Family
Anthropic has been shipping fast lately, a pace you can track on its news page. Claude Opus 5.5 and Claude Sonnet 5.5 both launched in September 2026, and Haiku 5.5 followed in early October. This comes on top of Anthropic’s push to train more engineers through programs like Claude Frontier Academy, so the company is clearly investing in both its models and the people who build on them.
Google has been moving at a similar speed. Our breakdown of what Gemini 4 Argon is covers a similar story from Google’s side: better performance per dollar, longer context windows, and models tuned for specific jobs rather than one model trying to do everything.
If you want the simplest way to think about it: Opus is for your hardest reasoning tasks, Sonnet is the balanced option most people use day to day, and Haiku is for speed and scale. Anthropic has committed to keeping Haiku 5.5 available until at least October 2027, so it is a safe model for developers to build around for at least the next year.
Common Questions
Is Claude Haiku 5.5 free to use?
No. It is a paid API model, priced per million tokens. Some apps that use Claude behind the scenes may offer their own free tiers, but the model itself is not free for developers to run.
Can I use Claude Haiku 5.5 right now?
Yes, if you have access to the Claude API, Claude Platform, Amazon Bedrock, Google Cloud, or Microsoft Foundry. It has replaced Haiku 4.5 as Anthropic’s current small model.
Is Haiku 5.5 better than Sonnet or Opus?
Not for complex reasoning. It is designed to be faster and far cheaper for simpler, high volume tasks, not to outperform the larger models on hard problems.
Why did the price drop so much?
Anthropic has been optimizing its smaller models for scale, since features like classification, data extraction, and routing run small models millions of times a day. A cheaper small model has a bigger impact on total AI costs than a cheaper large model does.
Final Takeaway
Claude Haiku 5.5 will not make headlines the way a flashy new chatbot feature does, but it is the kind of update that quietly makes everyday AI tools faster and cheaper to run. If you keep an eye on how your favorite apps update their AI features over the next few months, there is a decent chance a model like this one is working in the background, and now you know what to look for.
by admin | Oct 12, 2026 | AI Guides
If you have ever wished you could just tell a computer what you want done instead of clicking through five different apps to do it yourself, Google just moved a step closer to that. On October 8, 2026, at its “Gemini at Work” event, Google announced the Gemini AI agent, a single, universal agent for the workplace. Not a chatbot that answers questions. An agent that actually finishes the task.
This is a business-first launch, so you will not see it in your personal Gemini app tomorrow morning. But it is worth understanding now, because it shows where AI tools are heading for everyone, including people who will never touch an enterprise dashboard.
What the Gemini AI Agent Actually Does
According to Google’s own announcement, the new Gemini agent works from a single prompt box. Instead of typing out step by step instructions, you give it an objective, and it plans the work itself. It can pull in different skills and tools, connect to a company’s internal systems, and hand back a finished result inside the apps people already use, like Docs, Gmail, Slack, or a developer environment.
A few details stand out. The agent can run for hours or even days on a task, spin up temporary sub-agents to handle pieces of the job, and act almost like a coworker. It gets its own identity and an email address ending in agents.company.com, so you can loop it into a conversation the same way you would loop in a colleague. Its actions also show up in an audit trail, which matters for companies tracking who, or what, did what.
Google says the agent connects to tools like Confluence, Microsoft Office, Teams, Slack, Jira, Salesforce, ServiceNow, BigQuery, Databricks, Postgres, and Snowflake, plus any Model Context Protocol (MCP) server inside or outside a company’s network. That’s a long list, and it tells you this was built for real workflows, not a demo.
It Doesn’t Only Use Google’s Own Models
Here’s the part I found genuinely interesting. Google’s own blog post confirms the agent orchestrates across both the Gemini model family and Claude models from Anthropic, picking whichever model fits a task best to balance quality and cost. Google calls this “multi-model orchestration,” and it pairs it with smart routing and real-time spend caps so a company can set a hard limit on what a project is allowed to spend on AI before anything gets paused.
If you want to see how fast Google’s own models have been moving on the capability side, our breakdown of Gemini 4 Argon covers the reasoning model sitting behind a lot of this.
Why This Matters Even If You’re Not at a Big Company
This launch is aimed at enterprise customers right now, with names like PayPal, Shopify, Deutsche Bank, and The Home Depot already testing it. But the shift it represents, from “AI that answers” to “AI that does,” is the same shift showing up across the industry. Google’s own explainer on what an AI agent actually is is a good plain-English primer if the term still feels fuzzy.
From my own experience working on websites and online tools, the useful AI jump has never been the chatbot that writes a nice paragraph. It’s the tool that logs into the right system, pulls the right file, and finishes the boring multi-step task without me babysitting every click. That’s the gap Google is trying to close here, and it’s the same gap Claude and ChatGPT have been closing from their own direction.
Tip: if your workplace rolls out an AI agent like this, start by handing it one well-defined, low-risk task (a weekly report, a data pull, a routine email draft) before trusting it with anything that affects customers or money. Agents are only as good as the instructions and guardrails you give them.
What This Means If You’re Job Hunting or Already Employed
Google is rolling this out to businesses first and consumers later, so there’s no exact date yet for when an ordinary Gemini user gets the full agent experience. For now, the practical takeaway is simple: employers are going to expect more comfort with AI-assisted workflows, not less. If your job touches reporting, scheduling, customer support, or document work, it is worth getting ahead of this rather than waiting for it to show up on your desk. Our guide on how AI is changing future jobs and our roundup of useful AI tools for daily work and study are both good next stops. If you want to see how Google framed its last wave of workplace AI features before this one, we also covered the earlier Gemini Workspace integrations rollout.
Common Questions
Is Google’s Gemini agent available to regular users yet?
Not fully. Google says enterprise customers come first, with a consumer rollout planned later. No general public launch date has been announced.
Does this replace human jobs immediately?
No. Google frames it as a tool businesses assign specific tasks to, with humans setting the objectives and reviewing the audit trail. Early adopters are using it alongside existing teams, not instead of them.
Can the agent use AI models other than Google’s own?
Yes. Google confirmed it can orchestrate across Gemini models and Claude models from Anthropic, and plans to add more models later.
Do I need to learn anything new to prepare for this?
Not urgently, but it doesn’t hurt to get comfortable giving AI tools a goal instead of a rigid step-by-step script. That “objective, not instructions” mindset is becoming the norm across most major AI tools, not just Gemini.
Final Takeaway
Google didn’t just update a chatbot. It reorganized Gemini around the idea that AI should finish tasks, not just discuss them, and it’s betting enough on this that it’s letting the agent use a rival’s models to do it well. You don’t need to use it today to feel its effects. The expectations it sets for what “normal” AI help looks like at work are going to reach your job eventually, so it’s worth understanding the shift now rather than catching up later.
