You may be more or less prepared for the academics of college, but the other life stuff can be an eye-opener. College might be the first time you’re in charge of your own finances, and with new living situations, new jobs and new connections, you may also be expanding the amount of personal data you’re putting out into the world. If you could use a little help with budgeting, remembering passwords or making sure everything you do online is secure, here are the finance and security apps we’ve used, tested and ultimately recommend.

Mint

If you’re new to tracking finances, getting an overview of your banking, credit and loan accounts in one place can be helpful. Mint is a simple and free app that does just that. I tested it for our subscription guide and continue to use it. The interface is intuitive and it’s pretty good at correctly categorizing purchases. The main features, like transaction history, self-budgeting and goal-setting, are available free. For $5 per month, you can have Mint cancel subscriptions on your behalf and you won’t see as many ad links peppered throughout the app (though, I’ve never found the ads particularly distracting.)

YNAB

For help creating a more formal budget, a few Engadget staffers use YNAB (You Need A Budget) and we recommend it in our guide to student budgeting. It’s based around a theory that imposes four “rules” to improve your money management, and learning those principles now will benefit you long after graduation. The browser and mobile app interfaces are pretty easy to use, and YNAB has a ton of instructional content for newbies that can point you in the right direction when you’re first setting up expense categories, debt trackers and sinking funds. It’s usually $15 per month or $99 per year, but students who can prove they’re in school can get a year for free.

Goodbudget

Between loans, jobs and, if you’re lucky, scholarships and financial aid, a student’s “extra” money can be pretty limited. Goodbudget translates the envelope technique to an app format, earmarking your money for the things you need to pay for. By visualizing what you have and what you need, you can see when there’s room for stuff you want, like going out with friends or decorating your first apartment. Plenty of graphs and sliders help map out your situation, and Goodbudget also offers free online classes for those who want to get better with money (granted, that may be a hard sell when you’re already in school). The free version gives you twenty total envelopes, split between expenses and goals, and lets you add one bank account. For unlimited accounts and envelopes, the paid version is $8 per month or $70 per year.

Acorns investment

Say you indulge in an Iced Toasted Vanilla Oatmilk Shaken Espresso for $5.75. The Acorns investment app rounds up that last 25 cents and deposits it into an investment account, and over time, your money grows. By providing a simple app and recommending just a few different portfolios, Acorns takes some of the complexity out of investing. For students in particular, it’s also easier to invest a few cents here and there than larger chunks of cash when you’re already just trying to get by. The monthly plan defaults to $5 per month with an option of a $3 plan at sign up. Both come with a checking and a retirement savings account in addition to the investment features, so if you’re totally starting fresh, this could prove useful.

1Password

Our senior security reporter, Katie Malone, put 1Password at the top of Engadget’s guide to password managers. Like all services like this, 1Password one helps you create unique and complex credentials for every site you use, and then saves them securely so you don’t have to remember them all. It works across most platforms and even lets you share logins and credit card info with other people as needed, which will make it easier to access any family accounts you may need while in school. The security and encryption measures are top-notch, with a zero-knowledge policy that ensures the company doesn’t store your data, as well as a bug bounty program that rewards ethical hackers who discover any vulnerabilities.

Proton VPN

If you study in public places where the WiFi is suspect, a VPN can give you an extra layer of protection. It’s not a cure-all for online security woes, but VPNs do create a protected “tunnel” to keep out people who may otherwise have access to your data, like your internet service provider or hackers targeting public WiFi. Proton VPN is the best overall option not just because it’s easy to use. The Switzerland-based company also enforces a no-log policy and their open-source software continually stands up to independent audits. Unlike some VPNs, it didn’t tank our connection speeds in our tests, either. Proton goes for $10 per month to access servers in 65 countries, or you can get the free version with access to just three.

ProtonMail

Free email services are everywhere, but finding one that isn’t propped up by selling your habits and history to advertisers is almost impossible. And while you might get a school email address, a good personal email will serve you long after access to your alumni mail is discontinued. ProtonMail is focused on privacy: It uses end-to-end encryption, whereas a service like Gmail encrypts messages in transit only. Proton’s open-source encryption methods are independently audited, and since the service is supported by paid subscriptions and not advertising, the company has little incentive to snoop your info. Free plans give you one gigabyte of storage and allow for 150 emails per day, while a $12-per-month subscription grants 500GB of storage and removes email limits.

Signal

As a non-profit, there's no tech giant behind the wheel at Signal, which sets it apart from most other messaging services. A phone number is required for set up, but that’s about all the information Signal ever collects. It’s a favorite of journalists, protestors and people living in unstable territories, but students who realize their communications are no one else’s business will find the app useful, too. Texts, videos and images you send are end-to-end encrypted using open-source protocols, and you can even set messages to expire. Recent additions that enhance group chats may make Signal feel a little more like other messaging apps, but the core structure of the service will always be fundamentally more private than many competitors.

Noonlight

Staying safe in college extends beyond online safety, which is where apps like Noonlight come in. Tinder bought a stake in the app a few years ago to help people in the event of a date gone wrong. Within the app, you’ll find a giant white button that you press and hold in sketchy situations. As long as you hold the button, nothing happens. Let go of it, and unless you enter a secret pin to prove you’re safe, the police will be dispatched to your location. A timeline feature lets you add names and images when you’re meeting someone new. The safety network allows your friends and family to request check-ins and take action when they don’t hear from you. The free version includes all three of the features mentioned above, while the $5-per-month plan adds an iPhone widget and the ability to sync with rideshare apps.

At the Microsoft Inspire partner event today, the Windows maker announced pricing for its AI-infused Copilot for Microsoft 365. The suite of contextual artificial intelligence tools, the fruit of the company’s OpenAI partnership, will cost $30 per user for business accounts. In addition, the company is launching Bing Chat Enterprise, a privacy-focused version of the AI chatbot with greater security and peace of mind for handling sensitive business data.

Revealed in March, Microsoft 365 Copilot is the company’s vision of the future of work. The GPT-4-powered suite of tools lets you generate Office content using natural-language text prompts. For example, you can ask PowerPoint to create a presentation based on a Word document, generate a proposal from spreadsheet data or summarize emails and draft responses in Outlook — all by typing simple commands. “By grounding answers in business data like your documents, emails, calendar, chats, meetings and contacts, and combining them with your working context — the meeting you’re in now, the emails you’ve exchanged on a topic, the chats you had last week — Copilot delivers richer, more relevant and more actionable responses to your questions,” Frank X. Shaw, Microsoft’s Chief Communications Officer, wrote in an announcement today.

Microsoft began testing Copilot with a small group of select enterprise partners earlier this year but hasn’t yet announced when all business customers will gain access. However, announcing its pricing could mean that date is fast approaching. The $30 / mo. pricing will apply to Microsoft 365 E3, E5, Business Standard and Business Premium customers. The company still hasn’t announced Copilot consumer pricing or availability.

Meanwhile, Bing Chat Enterprise is Microsoft’s more security-minded variant of the popular AI chatbot that launched for consumers in February. “Since launching the new Bing in February, we’ve heard from many corporate customers who are excited to empower their organizations with powerful new AI tools but are concerned that their companies’ data will not be protected,” Shaw wrote. “That’s why today we’re announcing Bing Chat Enterprise, which gives organizations AI-powered chat for work with commercial data protection. What goes in — and comes out — remains protected, giving commercial customers managed access to better answers, greater efficiency and new ways to be creative.”

Bing Chat Enterprise begins rolling out today in a preview — at no additional cost — for Microsoft 365 E5, E3, Business Premium and Business Standard customers. In addition, the company says it will make the enterprise-focused chatbot available as a standalone $5 subscription “in the future.”

Tumblr could be the latest platform to borrow from TikTok’s playbook. The company is planning a major revamp of its platform that will bring algorithmic recommendations to users’ feeds, according to a memo published on the Tumblr Staff blog.

The memo is notably frank about the reasons for the upcoming changes and what it describes as Tumblr’s current shortcomings. “The underlying problem is that Tumblr is not easy to use,” the company writes. “Being a 15-year-old brand is tough because the brand carries the baggage of a person’s preconceived impressions of Tumblr.”

While Tumblr doesn’t provide exact details about new features, it offers some pretty big hints about what’s to come. The company says that one of its primary goals will be to “deliver great content each time the app is opened” and refers to its current “following” feed as “outdated.”

To address this, the Automattic-owned platform says it’s working to “improve our algorithmic ranking capabilities across all feeds” and “make it easier for users to understand where the vibrant communities on Tumblr are.” The company also notes that building more creator-friendly features, including improvements to the way replies and reblogs work, will also be key to attracting new users.

“Being a new creator on Tumblr can be intimidating, with a high likelihood of leaving or disappointment upon sharing creations without receiving engagement or feedback,” the company writes. “The lack of feedback stems from the outdated decision to only show content from followed blogs on the main dashboard feed (“Following”), perpetuating a cycle where popular blogs continue to gain more visibility at the expense of helping new creators.”

Taken together, the changes Tumblr is describing sound a lot like TikTok (or even Instagram): algorithmic recommendations in users’ primary feeds, creator-friendly features that encourage sharing, and more streamlined commenting and conversation tools. As a strategy, that all may sound pretty straightforward in 2023, when users increasingly expect these kinds of features from social platforms anyway. But considering Tumblr’s core interface hasn’t changed that much in its decade and a half of existence, the new direction could bring significant changes to the overall dynamics of the platform.

The coming redesign isn’t the only way Automatic has tried to breathe new life into the platform it acquired in 2019. The company has also experimented with subscriptions and other paid features, introducing Post+ in 2021, though there was some backlash against the feature from longtime users. More recently, the company began selling “completely useless” checkmarks to users soon after Elon Musk’s botched rollout of Twitter’s new paid verification.

When it comes to online merchandizing, nobody is bigger than Amazon. The same can be said for Walmart's utter domination of physical retail. But for a brief time in 2016, the two behemoths sought to get all up in each other's lanes. The resulting multi-year fracas would shake the world of commerce to its foundations with every above-board strategy and under-handed trick made available to crush the competition. In Winner Sells All, journalist Jason Del Rey recounts the business battles both between and within these titans of industry as both corporations sought to further entrench their market positions. In the excerpt below, we see some of said underhanded tricks.

Harper Collins

Excerpted from Winner Sells All: Amazon, Walmart and the Battle for Our Wallets by Jason Del Rey . Published by Harper Business. Copyright © 2023 by Jason Del Rey. All rights reserved.

In the late 2010s, the power and valuations that Amazon and other titans of the technology industry were accumulating incited a new movement in antitrust circles, catalyzed by a law school paper written by a then-unknown law student named Lina Khan. In her seminal paper, “Amazon’s Antitrust Paradox,” published in the Yale Law Journal, Khan argued that our interpretation of antitrust laws was outdated in light of a new digital economy, and there was a need to return to the days when merely having low prices or providing free services wasn’t enough to avoid scrutiny for anticompetitive behavior. 

“Amazon doesn’t just want to dominate markets; it wants to own the infrastructure that underpins those markets,” said Stacy Mitchell, the longtime critic of both Amazon and Walmart who runs a left-leaning think tank called the Institute for Local Self-Reliance (ILSR). “And that’s an order of magnitude difference of a monopoly ambition than Walmart’s.” Mitchell had spent many years agitating for the government to step in to slow down Walmart during its go-go Supercenter growth years and she is still clear today that she finds the company’s power problematic. 

But in her view, and that of many Big Tech critics in her circles, Amazon poses an altogether different threat to business competition. “It’s not just the retail platform, but it’s AWS [Amazon Web Services], it’s the logistics piece, it’s [Alexa] and being the interface for how we interact with the web, and all the devices and everything that are connected to the smart home,” she said. “It enables Amazon to favor its own goods and services in those markets, to levy a kind of tax on all the businesses that rely on that infrastructure, and to surveil all of that activity and use that intelligence to its own advantage.”

As the pressure from Washington, DC, increased, Amazon leaders were becoming heated. In one key annual meeting of Bezos’s senior leaders in early 2020, Jassy, the then-CEO of AWS, digested the content of a memo sitting in front of him. It laid out Amazon’s plans for messaging in response to accusations that it was too big or too powerful and engaged in anticompetitive behavior. As Bezos listened in by phone, Jassy pointedly asked those before him why the messaging didn’t argue that Walmart, and AWS rival Microsoft, should be investigated. Other top company officials tried to explain that each of those companies had already been scrutinized years ago and their time had passed. But Jassy’s reaction left a lasting impression on those in attendance.

“It was very clear from his comments that we shouldn’t let our foot off the gas,” someone in attendance told me years later. In subsequent years, especially in the part of the company that focused on so-called competition issues, “there wasn’t a day that Walmart didn’t come up.” The fact that Walmart, with more annual revenue than Amazon, was not being scrutinized by policy makers drove executives like Jassy crazy. It didn’t help when Amazon executives discovered that Walmart was indirectly funding a nonprofit front group called Free and Fair Markets, which was bombarding reporters and social media with anti-Amazon accusations. For some time, Amazon leaders suspected that a competitor, or group of competitors, was funding the operation but couldn’t prove it. One of Amazon’s longtime spokesmen, Drew Herdener, grew frustrated every time the group placed an op-ed or social media message that got traction.

“How does the press not know this is a front group?” he would lament. As a result, an Amazon communications staffer named Doug Stone spent upward of a year trying to help reporters uncover the group’s funders. Finally, in the fall of 2019, the Wall Street Journal pulled back the veil in an expose titled “A ‘Grass Roots’ Campaign to Take Down Amazon Is Funded by Amazon’s Biggest Rivals.” A Walmart spokesperson denied funding the group to the newspaper—the article had stated that Walmart used an intermediary to pass along funds to FFM, so the company’s defense might have been a matter of semantics— but said that Walmart “share[s] concerns about issues” that the group was publicizing.

Amazon's $1.7 billion acquisition of Roomba manufacturer iRobot is under scrutiny not only in the US, but also in Europe. The European Commission has revealed that it has opened an in-depth investigation into the purchase out of concerns that the merger would restrict competition for robotic vacuum cleaners. In particular, the commission is concerned that Amazon might prevent iRobot's rivals from selling their robot vacuums on its marketplace. Amazon might favor iRobot's products in both paid and unpaid search results or charge competing products more to advertise their offerings, the commission said. 

In addition, authorities are worried about the possibility of Amazon preventing iRobot rivals from accessing future Alexa APIs and from getting the "Works with Alexa" certification, since voice control with the assistant is one of the most important selling points for robot vacuums. Finally, the commission has raised concerns about Amazon obtaining access to iRobot users' data, which it believes could provide the e-commerce giant "an important advantage." Amazon could, for instance, use that information to better personalize and target advertisements. Bottom line is that the "transaction may raise barriers to entry and expansion for Amazon's competitors to the detriment of consumers."

The commission will now look into these concerns in cooperation with other competition authorities and has until November 15th to finalize its results. In the US, the Federal Trade Commission launched a probe into the acquisition last year after politicians from both parties asked it to oppose the purchase.

An Amazon spokesperson told Reuters that the company will "continue to work through the process with the European Commission and [is] focused on addressing its questions and any identified concerns at this stage." They also told the news organization that Amazon could "offer a company like iRobot the resources to accelerate innovation and invest in critical features while lowering prices for consumers."

OnePlus has unveiled the Nord 3, and it's a significant upgrade over the 2T — to the point where you might prefer it over comparable rivals. The new mid-range model jumps to a 6.7-inch, 120Hz 2,772 x 1,240 screen, promising a big boost over the 6.4-inch, 90Hz 1080p panel from its predecessor. You'll also get 80W wired fast charging the company claims can deliver a 60 percent charge in 15 minutes.

The OnePlus Nord 3 also makes the leap to a MediaTek Dimensity 9000 chip (versus the 2T's Dimensity 1300), and it now comes with up to 16GB of RAM instead of last year's 12GB. The battery is larger, too, at 5,000mAh versus 4,500mAh. The device's 50MP primary camera is the same as on the higher-end OnePlus 11. That won't beat the absolute best cameras on the market, but it's more compelling in this price class. You can also expect an 8MP ultra-wide camera, a 2MP macro sensor and a 16MP front cam.

Software support is stronger this time around. OnePlus now promises three years of major OS updates for the Nord 3 compared to just two for the 2T. You'll also receive four years of security fixes. That's not as extensive as for the OnePlus 11 or rivals like Samsung, but it still means your phone will likely remain up to date until you're ready to replace it.

OnePlus is taking Nord 3 pre-orders today, starting at €449 ($488) for 8GB of RAM and 128GB of storage. Pay €549 ($597) and you'll get 16GB of RAM alongside 256GB of storage. At the moment, though, it's only poised to reach 16 European Union countries as well as key markets like India. It's not surprising that OnePlus won't release the phone in the US (the company only offers its entry Nord N-series stateside), but it's also not coming to the UK as of this writing. A spokesperson toldTech Advisor last week that "logistical challenges" with customs prevented it from offering the Nord 3 and other phones online. You may still see a UK debut, but not in the immediate future.

Whether or not it's a good buy versus the Google Pixel 7a or Samsung Galaxy A54 depends on your priorities. OnePlus is offering some of the best display, camera hardware and memory in this category, but Google typically produces better real-world photos and offers a 'pure' Android interface with timelier upgrades. Samsung, meanwhile, offers better secondary cameras than OnePlus, more updates and tight integration with its ecosystem.

Niantic is joining the long list of companies laying off employees. The studio published an “organizational update” (after the internal memo was leaked toKotaku) announcing that it would let go of 230 employees. In addition, the company is shutting down its Los Angeles studio and canceling two licensed games: NBA All-World and Marvel: World of Heroes. The only good news for fans is that Pokémon Go, still its flagship product, will live to fight another day.

“I have made the decision to narrow our focus for mobile game investments, concentrating on first party games that most strongly embody our core values of location and local social communities,” wrote CEO John Hanke. “The mobile gaming market is very mature and only the best and most differentiated titles have a chance to succeed. We also want to increase our focus on building for the emerging class of MR devices and future AR glasses.”

Although nobody likes hearing about layoffs, Hanke’s letter seems forthcoming and candid about the company’s challenges and the mistakes he and the leadership team made. He attributes the downsizing to the studio growing its expenses faster than its revenue. “In the wake of the revenue surge we saw during Covid, we grew our headcount and related expenses in order to pursue growth more aggressively, expanding existing game teams, our AR platform work, new game projects and roles that support our products and our employees. Post Covid, our revenue returned to pre-Covid levels and new projects in games and platform have not delivered revenues commensurate with those investments.” The CEO expects the reductions to “bring expenses and revenue back into line” without shuttering its most valuable property.

On that topic, Hanke said, “The top priority is to keep Pokémon Go healthy and growing as a forever game. While we made some adjustments to the Pokémon Go team, our investment in the product and team continues to grow.” The phone-based AR game launched in the summer of 2016 and instantly became a viral sensation; it also enjoyed a resurgence in popularity during pandemic lockdowns as homebound gamers used the explorative game as an excuse to connect with a world beyond their overly familiar four walls. Just last year, the studio launched a social app for the game that lets players organize and chat.

The letter adds that the AR market “developing more slowly than anticipated” was another factor. Niantic’s games would be ideal for on-the-go AR, which the industry eventually expects to materialize as smart glasses that can pass for standard prescription specs. However, as Apple’s Vision Pro made clear, that future is still likely a ways off. Today’s wearable AR, also including the Meta Quest Pro, is home-based gear designed for work and entertainment in the comfort of your home or office. Although adaptations of Pokémon Go types of experiences may work there to some degree, the company’s trademark approach is tailor-made for AR that isn’t yet available. Niantic may well end up waiting five to 10 years to see consumer-friendly versions of that type of augmented reality — and apparently, that required some reconfiguring.

Unless you switch to chronological timeline on Facebook, the things you see on your timeline could seem pretty arbitrary at times. Now, Meta has given us pretty comprehensive look at how its AI systems decide on the posts that appear on our Facebook and Instagram feeds. The social networking giant has released 22 card systems for the platforms that explain how its AI systems rank content for its Feed, Stories and Reels. Each card explains how a certain aspect of Meta's platform works — for instance, the company explains that for Facebook Feeds, its AI system starts by gathering all potential posts by friends and Pages you follow. 

After that, the system considers various input signals, such as who created the post, how you interacted with them and how many of your friends have liked the post, if any. All those aspects will help the AI rank posts based on which you'll find the most relevant and valuable and then show them on your feed in order of the scores they get. The card systems also show how you can customize what you see on the platforms and link to instructions on, say, how to unfollow a person or a group and how to click "show more" or "show less" on a post to indicate if you like or dislike a particular subject. 

In a post announcing the cards' release, Nick Clegg, Meta's President of Global Affairs, listed the tools you can use to personalize your experience on the company's platforms. In addition to the two aforementioned tools, Facebook and Instagram have centralized menus with customization controls, called Feed Preferences and Suggested Content Control Center, respectively. Clegg also said that Meta is testing a new feature on Instagram allows you to indicate whether you're "Interested" in a recommended reel in the Reels tab, so the app can show you more similar content. At the moment, you can only indicate whether you're "Not Interested" in a specific reel. In addition, he said Meta is working on making the "show more" and "show less" feature more prominent. 

Meta has lifted the lid on how its algorithms work before the European Union starts enacting the Digital Markets Act in 2024. The new law will require online services like Facebook and Instagram to be more transparent on the technologies behind their recommendation algorithms. It will also require platforms to offer chronological feeds and will ban ads targeting you based on your religion, sexual orientation, ethnicity or political affiliation. Earlier this year, Meta also rolled out a new version of its "Why am I seeing this ad?" tool to provide more transparency on how your activities power its ad-matching software.

“Learn to code.” That three-word pejorative is perpetually on the lips and at the fingertips of internet trolls and tech bros whenever media layoffs are announced. A useless sentiment in its own right, but with the recent advent of code generating AIs, knowing the ins and outs of a programming language like Python could soon be about as useful as knowing how to fluently speak a dead language like Sanskrit. In fact, these genAIs are already helping professional software developers code faster and more effectively by handling much of the programming grunt work.

How coding works

Two of today’s most widely distributed and written coding languages are Java and Python. The former almost single handedly revolutionized cross-platform operation when it was released in the mid-’90s and now drives “everything from smartcards to space vehicles,” according to Java Magazine in 2020 — not to mention Wikipedia’s search function and all of Minecraft. The latter actually predates Java by a few years and serves as the code basis for many modern apps like Dropbox, Spotify and Instagram.

They differ significantly in their operation in that Java needs to be compiled (having its human-readable code translated into computer-executable machine code) before it can run, while Python is an interpreted language which means that its human code is converted into machine code line-by-line as the program executes, enabling it to run without first being compiled. The interpretation method allows code to be more easily written for multiple platforms while compiled code tends to be focused to a specific processor type. Regardless of how they run, the actual code-writing process is nearly identical between the two: somebody has to sit down, crack open a text editor or Integrated Development Environment (IDE) and actually write out all those lines of instruction. And up until recently, that somebody typically was a human.

The “classical programming” writing process of today isn’t that different from the process those of ENIAC, with a software engineer taking a problem, breaking it down into a series of sub-problems, writing code to solve each of those sub-problems in order, and then repeatedly debugging and recompiling the code until it runs. “Automatic programming,” on the other hand, removes the programmer by a degree of separation. Instead of a human writing each line of code individually, the person creates a high-level abstraction of the task for the computer to then generate low level code to address. This differs from “interactive” programming, which allows you to code a program while it is already running.

Today’s conversational AI coding systems, like what we see in Github’s Copilot or OpenAI’s ChatGPT, remove the programmer even further by hiding the coding process behind a veneer of natural language. The programmer tells the AI what they want programmed and how, and the machine can automatically generate the required code.

Building the tools to build the tools allowing any tool to build tools

Among the first of this new breed of conversational coding AIs was Codex, which was developed by OpenAI and released in late 2021. OpenAI had already implemented GPT-3 (precursor to GPT-3.5 that powers BingChat public) by this point, the large language model remarkably adept at mimicking human speech and writing after being trained on billions of words from the public web. The company then fine-tuned that model using 100-plus gigabytes of GitHub data to create Codex. It is capable of generating code in 12 different languages and can translate existing programs between them.

Codex is adept at generating small, simple or repeatable assets, like “a big red button that briefly shakes the screen when clicked” or regular functions like the email address validator on a Google Web Form. But no matter how prolific your prose, you won’t be using it for complex projects like coding a server-side load balancing program — it’s just too complicated an ask.

Google’s DeepMind developed AlphaCode specifically to address such challenges. Like Codex, AlphaCode was first trained on multiple gigabytes of existing GitHub code archives, but was then fed thousands of coding challenges pulled from online programming competitions, like figuring out how many binary strings with a given length don’t contain consecutive zeroes.

To do this, AlphaCode will generate as many as a million code candidates, then reject all but the top 1 percent to pass its test cases. The system will then group the remaining programs based on the similarity of their outputs and sequentially test them until it finds a candidate that successfully solves the given problem. Per a 2022 study published in Science, AlphaCode managed to correctly answer those challenge questions 34 percent of the time (compared to Codex’s single-digit success on the same benchmarks, that’s not bad). DeepMind even entered AlphaCode in a 5,000-competitor online programming contest, where it surpassed nearly 46 percent of the human competitors.

Now even the AI has notes

Just as GPT-3.5 serves as a foundational model for ChatGPT, Codex serves as the basis for GitHub’s Copilot AI. Trained on billions of lines of code assembled from the public web, Copilot offers cloud-based AI-assisted coding autocomplete features through a subscription plugin for the Visual Studio Code, Visual Studio, Neovim, and JetBrains integrated development environments (IDEs).

Initially released as a developer’s preview in June of 2021, Copilot was among the very first coding capable AIs to reach the market. More than a million devs have leveraged the system in the two years since, GitHub's VP of Product Ryan J Salva, told Engadget during a recent interview. With Copilot, users can generate runnable code from natural language text inputs as well as autocomplete commonly repeated code sections and programming functions.

Salva notes that prior to Copilot’s release, GitHub’s previous machine-generated coding suggestions were only accepted by users 14 - 17 percent of the time, “which is fine. It means it was helping developers along.” In the two years since Copilot’s debut, that figure has grown to 35 percent, “and that's netting out to just under half of the amount of code being written [on GitHub] — 46 percent by AI to be exact.”

“[It’s] not a matter of just percentage of code written,” Salva clarified. “It's really about the productivity, the focus, the satisfaction of the developers who are creating.”

As with the outputs of natural language generators like ChatGPT, the code coming from Copilot is largely legible, but like any large language model trained on the open internet, GitHub made sure to incorporate additional safeguards against the system unintentionally producing exploitable code.

“Between when the model produces a suggestion and when that suggestion is presented to the developer,” Salva said, “we at runtime perform … a code quality analysis for the developer, looking for common errors or vulnerabilities in the code like cross-site scripting or path injection.”

That auditing step is meant to improve the quality of recommended code over time rather than monitor or police what the code might be used for. Copilot can help developers create the code that makes up malware, the system won’t prevent it. “We've taken the position that Copilot is there as a tool to help developers produce code,” Salva said, pointing to the numerous White Hat applications for such a system. “Putting a tool like Copilot in their hands … makes them more capable security researchers,” he continued.

As the technology continues to develop, Salva sees generative AI coding to expand far beyond its current technological bounds. That includes “taking a big bet” on conversational AI. “We also see AI-assisted development really percolating up into other parts of the software development life cycle,” he said, like using AI to autonomously repair a CI/CD build errors, patch security vulnerabilities, or have the AI review human-written code.

“Just as we use compilers to produce machine-level code today, I do think they'll eventually get to another layer of abstraction with AI that allows developers to express themselves in a different language,” Salva said. “Maybe it's natural language like English or French, or Korean. And that then gets ‘compiled down’ to something that the machines can understand,” freeing up engineers and developers to focus on the overall growth of the project rather than the nuts and bolts of its construction.

From coders to gabbers

With human decision-making still firmly wedged within the AI programming loop, at least for now, we have little to fear from having software writing software. As Salva noted, computers already do this to a degree when compiling code, and digital gray goos have yet to take over because of it. Instead, the most immediate challenges facing programming AI mirror those of generative AI in general: inherent biases skewing training data, model outputs that violate copyright, and concerns surrounding user data privacy when it comes to training large language models.

GitHub is far from alone in its efforts to build an AI programming buddy. OpenAI’s ChatGPT is capable of generating code — as are the already countless indie variants being built atop the GPT platform. So too is Amazon’s AWS CodeWhisperer system, which provides much of the same autocomplete functionality as Copilot, but optimized for use within the AWS framework. After multiple requests from users, Google incorporated code generation and debugging capabilities into Bard this past April as well, ahead of its ecosystem-wide pivot to embrace AI at I/O 2023 and the release of Codey, Alphabet’s answer to Copilot. We can’t be sure yet what generative coding systems will eventually become or how it might impact the tech industry — we could be looking at the earliest iterations of a transformative democratizing technology, or it could be Clippy for a new generation.

As the unromantic name implies, TP-Link’s Deco XE75 AXE5400 is one of the company’s many, many routers. A trio of plain cylinders standing 6.7-inches tall, they mirror the lack of excitement in its name, but as the saying goes, let’s not judge a book by its deeply unsexy cover. Hiding inside is a mesh of extreme quality that, despite a few rough edges, offers a great mix of power and affordability. If you’re confident in your ability to work your way through an admin panel or two, then this might be the mesh for you.

Hardware

I’ve already mentioned that the XE75 comprises a series of nondescript-looking cylinders which aren’t the prettiest mesh units. They certainly look like WiFi equipment, unlike many of the others on the market, which look like paperweights designed by Henry Moore. Unless you’re living in a minimalist’s paradise, they’ll hopefully blend into your decor with no fuss.

On the back of each node are three gigabit ethernet ports, one of which you’ll need to hook the primary mode up to your modem. You can also use the ports to run ethernet backhaul, if your house is (or will be) suitably-equipped. In 2023, when it can sometimes feel like wired ethernet is becoming a niche proposition, having three ports per node feels like luxury. And I don’t think many users will gripe over a lack of a 2.5 Gbps port which is a bit excessive, even these days.

One of the first choices you’ll need to make is how you’ll use the 6GHz band, which is reserved as backhaul by default. You can leave it like this or, if you have a plethora of WiFi 6-enabled devices already, you can run it dynamically. Like I said in the mesh WiFi buyer’s guide, using the 6GHz band for backhaul makes sense for now since so few phones, laptops and tablets can access that band directly at the moment.

TP-Link says its “AI-driven mesh” will learn which devices get faster speed from which nodes and prioritize those connections accordingly. Once each of your devices is connected to the mesh, you should hope to see your speed and reliability improve as it learns your usage.

Installation

Deco XE75 was easy to get set up, taking just 16 minutes from when I pulled the plastic film from the box to finish. Download the Deco companion app, set up a TP-Link account and you’ll then be guided through the short setup process. You just need to plug the first node in, tell the app if you have any ISP-specific needs, give your network a name and password and you’re off at the races.

Once done, the app will tell you to turn on the other nodes in the set and wait as they connect to the existing mesh. You’ll also need to assign each node a name based on their locations in your home. My phone pinged several times in quick succession after this as all the gadgets in my home joined the network. The degree of seamlessness and ease of setup was more or less perfect.

That is, except for one minor real-world annoyance that I feel compelled to flag — the length of the power cables for the UK models I’ve been testing. A lot of mesh products ask users to place nodes in visible, prominent locations, rather than hidden behind furniture, to avoid interference. That’s harder to do, however, if your power cable is just 57 inches long, a small but noticeable bit shorter than the other products I’m testing. To the point where I couldn’t put one module in my usual location at the top of a bookcase because the cable didn’t stretch that far. In this case, that module had to go on my top shelf, rather than above it. It’s a specific and possibly niche complaint but worth mentioning in case you have an unforgiving room layout.

Performance

Photo by Daniel Cooper / Engadget

After setting up the modules in the usual places in my home, I found that it took about 10 minutes for the connection to stabilize. I had to run a firmware update that caused things to drop out for a further five minutes afterward, but after that, the connection was very stable. A lot of more affordable 6E routers use the 6GHz band as backhaul, and this worked well in my home.

Close to the primary node, speeds hit an average of 270 Mbps down, and in my office two floors away, I was still getting 260 Mbps. Even in my back bedroom with its dreaded signal dead spot, SpeedTest download benchmarks fell to around 220 Mbps. Ping times were similarly consistent, to the point where I reckon a two-point mesh might have sufficed.

The Deco app very clearly shows your network topography, enabling you to quickly see which devices connect to which node. What surprised me is that the hardware in my office preferred the primary node rather than the one that was nearer. I suspect, over time, those connections would shift, but the fact I saw such good performance despite being two floors away was great.

App and controls

The Deco app lays everything out in a friendly, easy-to-parse manner that shouldn’t deter novice users from upgrading. The home screen shows you the network topography, and what devices are connected to each node by default, helping you to feel in control of what’s going on.

If there’s a downside, it’s that the level of control available to you inside each submenu isn’t that deep. View your WiFi settings, for instance, and you’ll be able to change your network name and password or share those details to someone else. But the only other option is to decide if the 6GHz network is used for dedicated backhaul or if you can share it with devices on the network. You can activate a Guest Network on the homescreen, letting you set one up with one press, although I’m less of a fan that it’s password-free by default.

You can scroll the list of what’s connected to the network to see its signal strength, as well as how much data it’s up-and-down-loading at the time. Each device can be assigned to a family member for parental controls, and you can single out a unit for priority on the network. The one downside to this is that TP-Link really struggles to identify each piece of equipment on your network compared to, say, a product from Netgear. So many units were named “iot_device” in the list, that you’ll probably need to take the time to rename them all manually.

In the More sub-menu, you can run tests to optimize your network, set up an IPv6 connection, as well as tweak IP settings. One feature I appreciated was the choice to get a push notification every time a new device joins the network, which appeals to my paranoia.

You can also access your settings through a browser-based client but, as far as I could see, the only difference is it lets you force a firmware upgrade with a local file rather than handling the system online. That’s a fairly niche use, though.

Additional features

Deco does offer smart home integrations, but it’s limited to TP-Link’s own gear and Philips Hue. The only other thing that the company offers is Homeshield, which offers a suite of security features to help keep your WiFi secure. The free Basic tier will scan your network for security threats and offers “robust” parental controls. That includes the ability to block specific websites, set daily usage limits and time-out zones to stop your kids accessing the internet in the middle of the night.

You can also activate content filtering, which will lock down swathes of the web that TP-Link deems unsuitable. That includes Adult Content, Gambling, and Download sites, amongst many others. More problematically, you can block access to sites offering sex and relationship education information, which feels like TP-Link is enabling more harm than good there.

What I will say, despite my objections, is that the suite of options available for free here is a very good mix. Plenty of companies have taken to putting even the most basic parental controls, like time limits and access control, behind their paywalls. The fact the essentials are available here, for free, means the company gets plenty of extra points here.

TP-Link also offers a paid version of Homeshield, which includes more protection against hacks, greater data about what websites users are visiting. This, it says, will guard against “teenager internet addiction (sic),” “IoT Devices Attacks (sic)” and “Cyber Virus Intrusion (sic).” Homeshield Pro costs $5.99 a month, or $55 for a year, although I’m not sure I see enough value in it to encourage anyone to sign up for the extras on offer.

Wrap-up

There’s no single glitzy, attention-grabbing feature that makes the Deco XE75 a must-buy, but what puts it ahead of the competition is its brawn. Nestled inside those cylinders is powerful, reliable hardware that generates a fast and far-reaching network which is reason enough to pick up TP-Link’s system. The app and services offered at no additional cost, squarely tick the “good enough” box.

The one thing the Deco XE75 lacks is polish, both in its app and its services. I wouldn’t suggest this to anyone who would freeze up at the very thought of having to make a decision about something like a backhaul channel. But, if you are prepared to make the effort, then this is certainly the mesh WiFi system for you. It’s fast, reliable, fairly easy to use and I reckon the (cheaper) two-pack will cover all but the biggest of homes.