Skip to main content
  1. Home
  2. Branded Content
  3. Computing
  4. Emerging Tech
  5. Smart Home
  6. Features

Major Upgrade: Skywork Deep Research Agent v2 delivers a smarter, more efficient experience for AI Office and research

Add as a preferred source on Google
skywork deep research agent v2 feature
Image used with permission by copyright holder

While OpenAI’s ChatGPT may be the launching point for most people’s AI searches these days, it isn’t a be all, end all tool. And, despite Canvas mode, it certainly isn’t an office suite replacement tool. And that’s where emerging AI leaders are coming, filling in the gaps left by more open-ended tools. Just take Skywork, a forward-looking platform that’s taking on the AI Office space, helping create tools for researchers of all levels, make expert-tier spreadsheets, and PPT slides in a snap of the fingers. And, it’s getting better all the time, with the new Skywork Deep Research Agent v2 being revealed just recently, a new engine specifically designed for high-density documents, PPTs, and spreadsheets.

Here, we take a look at what Skywork has to offer now, as well as what it can bring into the future.

An easy research modeler

The Skywork site has tons of examples of its work that you can “replay” to see exactly how things go. This example report shows the AI’s steps in creating a sample report.

There are a few important details here that you shouldn’t miss and that ultimately make Skywork different. The first of these is the collaborative nature of the output. This isn’t a one shot, single prompt research paper. Instead, the system creates a list of requirements that you might have for it, letting you check what to add and what to skip, giving you ultimate control over the project. Likewise, a pre-writing to-do list is populated so you can tweak the requirements to favor what you need. Next is the consistent calling of MCP tools, which are the types of tools that AI is best suited to use autonomously at this time, making for an all around painless experience.

The results create a promising skeleton

And, what’s created certainly has the style, outline, and information needed for a report that can service an individual’s needs instantly and possibly head to publication with a guided hand. Formatting and some stylistic choices will certainly be helped by human editing, but that’s to be expected as each person and journal has different requirements.

Facts for the example report we reviewed earlier are made from MCP-fueled internet research, but it is easy enough to imagine inputting your own hard-earned data. And that’s where this will all shine. Researchers will do research, and Skywork will interpret it and present it with minimal help. This software doesn’t eliminate the researcher, but instead frees them from the burden of writing out their findings; an especially important service since not all researchers are adept in the near-universal standard of English language science reporting. Furthermore, it has achieved SOTA-level benchmarks in BrowseComp and GAIA Test, showing an ability to deliver results at the same level you do. This is a suite that works like an office suite to create your document, not replace the research and human thought needed to get there.

As Skywork, and AI in general terms, evolve, so will the ways we interact with written media, spreadsheets, and the old office suite altogether. Currently, Skywork serves as an excellent planner to get your research to the first draft stage quickly. Next year, who knows?

John Alexander
Former Digital Trends Contributor
John Alexander is a former ESL teacher, current writer and internet addict, and lacks the wisdom to know what the future…
This award-winning phone thermal camera is 10% off for Halloween, saving you $107
Thermal Master P3 thermal camera drops with code THERMCAM10 (save $107 total): 512x384 IR super resolution, an adjustable-focus lens, and 2026 Tom’s Guide AI Award winner.
Man using thermal camera attachment with phone

A phone thermal camera is one of those tools you don’t think you need until the first time it saves you from an expensive problem, and the Thermal Master P3 just got easier to try. For Halloween, it’s discounted with the code THERMCAM10, a 10% cut that works out to $107 off the total. It’s also fresh off a win at the 2026 Tom’s Guide AI Awards for best phone thermal camera, which is the kind of third-party nod that separates a serious tool from the cheap clip-on sensors that flood this category.

Thermal Master Halloween sale

Read more
Beyond Tracking: How AI Fitness Companions Are Changing Solo Training
Weapon, Toy, Green

Imagine you’re eight reps into a set of goblet squats in your living room. Your legs start to burn, your core is braced, and your focus remains sharply on finishing the set. But as fatigue sets in, an age-old uncertainty creeps in: Are my hips dropping below parallel, or am I cutting depth short? Is my lower back rounding under load? Am I doing this right?

For much of the past decade, mainstream consumer fitness technology has focused on quantification—tracking metrics such as steps, heart rate, workout duration, and calorie estimates. Those metrics can be useful for understanding activity and exertion, but they reveal far less about how a movement was actually performed: whether form changed as fatigue set in, whether squat depth shortened, or whether repeated movements stayed consistent from one rep to the next. 

Read more
Before a dash cam hits the road, it has to survive the lab
Technician standing beside orange testing chambers.

A dash cam has a fairly uneventful life once it reaches a car. It sits behind the windshield, records the road ahead and, ideally, goes unnoticed until the moment its footage is needed. Getting it ready for that job can be considerably less quiet.

Inside Wolfbox’s Central Laboratory, cameras and other automotive products are exposed to extreme temperatures, vibration, mechanical shock, water and prolonged operation. Known for its off-road dash cams and automotive electronics, Wolfbox uses the facility to recreate the difficult conditions its products may encounter before they reach drivers. The idea is simple enough. If a product is expected to work when conditions are difficult, those conditions need to be recreated before a driver encounters them.

Read more