FlashFeed

Full text

Qualcomm is finally getting serious about AI infrastructure, but its push into the datacenter hinges on the success of an ambitious near-memory compute architecture designed to deliver better inference economics than today's GPUs. Announced during its 2026 investor day last week, the tech will see Qualcomm stack layer upon layer of DRAM on top of its XPUs to form a single unified compute and memory module it's calling high-bandwidth compute (HBC). “We offer all of the performance advantages of SRAM, but with the density and the memory capacity that HBM (high-bandwidth memory) stacks offer,” Tony Pialis, Qualcomm’s EVP of datacenter, claimed during last week's investor presentation. This technology is set to launch next year as part of Qualcomm’s AI250-series of Dragonfly rack systems, and marks a distinct shift in Qualcomm’s AI infrastructure strategy. The handset giant is no stranger to AI accelerators. Essentially every Snapdragon processor sold today ships with an NPU on board. But in the datacenter, the company has struggled to garner the same excitement as Nvidia, AMD, and even startups like Cerebras. Compared to the big two’s GPUs, Qualcomm’s AI-series accelerators haven’t compared that favorably, but that could soon change as the company looks to make its mark on the datacenter. With the AI250, the SoC maker is claiming 768 GB of memory capacity and up to 133 TB/s of effective memory bandwidth per card. For reference, Nvidia's Groq 3 LPUs offer just 500 MB of SRAM and 150 TB/s of bandwidth. If that seems too good to be true, that’s because it is. Qualcomm is leaning heavily on the word “effective.” We know that because for the AI200-based Dragonfly systems rolling out this year, they claimed 414 TB/s of “effective” memory bandwidth across all 56 chips. On its face, that seems more realistic, but actually achieving that with 8800 MT/s LPDDR5x alone would require a 6,720-bit-wide bus, which it almost certainly does not possess. Qualcomm insists that this is the "pure physical bandwidth of the LPDDR interface," but declined to offer any specifics as to how it's somehow managed to achieve what Nvidia needed eight HBM3e stacks to do. In any case, according to Qualcomm’s marketing materials, with the move to HBC, the AI250 will offer 18x the effective bandwidth of the AI200, while the forthcoming AI300 will deliver 54x the bandwidth. Given the context, these seem like outlandish claims, but these "effective" multipliers are really a feature of Qualcomm's HBC architecture. Unpacking high-bandwidth compute Amplifying “effective” bandwidth isn’t the only party trick from these HBC-based accelerators. Qualcomm claims that by moving some of the XPU’s compute under the DRAM, it can significantly reduce the amount of power its chips consume. On a conventional datacenter GPU, data is rapidly shuffled between HBM and the compute dies. Even using advanced packaging technologies like TSMC’s CoWoS, the power required to move this data back and forth is significant. By stacking the DRAM directly on top of some of the logic and connecting them using through-silicon vias (TSVs), the path from compute to memory is shortened considerably. "Imagine working in the same building that you live in so you only travel up and down," Pialis said. "What does that mean for the highways and the roads that connect the suburbs to the city? Guess what? The roads are clear. The value this brings to the industry is lower power consumption, less heat, and that expensive road of silicon interposer that HBM solutions use is no longer needed." Performing bandwidth-bound operations on the base die also has the benefit of reducing the amount of data that needs to be shuttled to and from the HBC to the SoC. In effect, memory bandwidth is amplified. This is why Qualcomm is using “effective bandwidth" so liberally. Compared to doing all of that work on a conventional GPU or XPU with distinct HBM and compute dies, the effective bandwidth would be significantly higher, which also achieves better density than SRAM-only designs, like Nvidia’s LPUs or Cerebras’ dinner plate sized accelerators. With that said, Qualcomm probably won’t be running its entire AI software stack on HBC. Higher memory bandwidth primarily benefits decode, when the entirety of the model’s active weights are streamed autoregressively from memory one token after another. Decode isn’t particularly compute-intensive. As such, doing decode partially or entirely in HBC starts to make a lot of sense because it also avoids the thermal constraints associated with burying the compute under multiple layers of DRAM. Qualcomm tells us that the AI250 can be used as a standalone AI accelerator, but notes it is heavily optimized around addressing bandwidth bottlenecks. So, in addition to being a dedicated inference chip, it can be used in disaggregated inference architectures that use GPUs or other Qualcomm parts for prompt processing and the AI250 to speed up memory intensive decode operations. Peak FLOPS are notably missing from Qualcomm’s AI250 disclosures — the company declined to share specifics upon our request. Is HBC actually a competitive advantage? While Qualcomm is early among chip designers to make a fuss about near-memory or HBC, it’s not the first, nor is the technology beyond the means of Nvidia or AMD. In fact, both Nvidia and AMD are rumored to be working with HBM suppliers and TSMC to develop custom base dies to boost the performance of their next-gen chips, though it's still not clear how much, if any, compute has been integrated into them. Qualcomm tells us its HBC "uses LPDDR memory in a purpose-built near-memory computing architecture that combines compute and highly-accelerated memory bandwidth within a 3D-stacked silicon design. While both HBC and HBM use stacked-memory concepts, HBC is a distinct architecture designed to address AI’s data-movement bottleneck by bringing compute and memory closer together, increasing memory bandwidth efficiency and improving energy efficiency for AI inference workloads. HBM has more stacks of DRAM, uses 2.5D interposer to route more wires, and does not do computing in the base logic die." AI chip startup d-Matrix is also developing accelerators that will use 3D stacked DRAM to extend their in-memory compute capabilities. The underlying technology described by Pialis may not be as unique as Qualcomm would like investors to believe, but it shows the company hasn’t missed the boat. However, Qualcomm’s ability to work with Nvidia and AMD may end up doing more to sell customers on its tech than anything. As we previously wrote, in a disaggregated AI world, Nvidia can be both a friend and an enemy. Qualcomm finds its Mojo In addition to teasing its upcoming AI250 and AI300 accelerators, Qualcomm’s investor day also coincided with the acquisition of AI software startup Modular. Modular was founded by Tim Davis and Chris Lattner, the latter of whom you may recognize as the creator of LLVM, Clang, the Swift programming language, and the multi-level intermediate representation (MLIR) compiler infrastructure. At Modular, Lattner and crew developed Mojo, a low-level programming interface for GPUs, which offered a high-performance alternative to Nvidia’s CUDA or AMD’s HIP and ROCm stacks. The big idea is that users should be able to write highly performant AI apps that’ll run regardless of the underlying hardware. For Qualcomm, Mojo presents an opportunity to sidestep the CUDA moat, which has dogged AMD for so long. With Mojo, Qualcomm’s customers won’t need to choose one platform; they can develop their apps and run them on whatever compute is handy at the time. It’s not all or nothing either. Modular should help to support heterogeneous deployments similar to what Nvidia is doing with Groq’s LPU tech, where GPUs might be used for prefill and AI250s are used for decode in whatever ratio makes the most sense for that specific application. However, the acquisition doesn’t just buy Qualcomm a vendor-neutral programming model. The folks buying these systems are primarily concerned with one AI workload in particular: LLM model serving. For this, Modular developed a serving platform called Max. Max is a bit like SGLang or vLLM in that it’ll run interchangeably on AMD or Nvidia hardware, but because it’s built atop Mojo, it, at least in theory, shouldn’t require nearly as much hand tuning. The offering should help Qualcomm compete in a landscape where software has become even more important than the hardware it runs on, if it manages to close the acquisition this year without regulators stepping in. In any case, we won’t have to wait much longer to see the HBC in action. After launching its AI200-series racks later this year, Qualcomm plans to push its first-gen HBC-based AI250 out the door beginning in 2027, while its second-gen HBC platform is slated for 2028. While you wait, why not read up on Qualcomm’s new datacenter CPU, which we explored in more detail last week. ®

Full text

TL;DR

Samsung is starting a new campaign on Instagram to tease its upcoming launches.

The early videos it’s previewed sure appear to hint at the new, wide Galaxy Z Fold8.

So far, though, we haven’t seen any new teases about the actual launch date.

Samsung’s got some major new summer hardware launches coming up… right? At least, we’ve sure covered a ton of leaks and rumors about foldables like the Galaxy Z Flip 8, Fold 8, and Fold 8 Ultra , as well as the Galaxy Watch 9 and Watch Ultra 2 wearables, to say nothing of the Galaxy Ring 2 . And while we’ve heard rumblings about a possible July 22 Unpacked event , that’s only a few weeks away now, and we have yet to see Samsung sending out any invitations. The company is still being a little hush-hush about its ultimate plans, but with July just about to start, we’re learning about a new social media campaign to hint at what’s coming.

Starting tomorrow, July 1, Samsung is clearing history on its Instagram account , and will start dropping clues designed to inspire curiosity “built around familiar objects, each transformed in unexpected ways.” The company explains:

Full text

If there was any doubt left that Apple TV is the premiere destination for streaming science fiction, Silo Season 3 put it rest. The new season premieres on Friday, July 3, but critics who have gotten an early look are already raving about the new episodes. On Rotten Tomatoes , the season already has a 100% Fresh rating, with nine reviews from trusted critics accounted for so far. This is a great sign for Silo fans, but it also says a lot about Apple TV's investment in prestige sci-fi dramas in general. The streamer has built a brand on this genre, and unlike its competitors, Apple has stayed the course rather than relying on big hits to carry the whole brand. The result is a rising tide that lifts all ships.

Full text

Spy photographers have snapped the new ID. Tiguan undergoing testing.

It’s an extensively updated version of the ID.4 and features a new design.

Model will transition to the MEB+ platform and likely feature new motors.

The Volkswagen ID.4 went out of production in Chattanooga, Tennessee in April as the plant is embracing “higher‑volume products that meet market demand.” While the electric crossover is dead in America, the automaker has previously said existing inventory should be enough to “support customer demand into 2027.”

var adpushup = window.adpushup = window.adpushup || {que:[]};

adpushup.que.push(function() {

if (adpushup.config.platform !== "DESKTOP"){

adpushup.triggerAd("0f7e3106-c4d6-4db4-8135-c508879a76f8");

} else {

adpushup.triggerAd("82503191-e1d1-435a-874f-9c78a2a54a2f");

}

});

After that point, a “future version” will arrive in America. It’s known as the ID. Tiguan and spy photographers recently snapped a prototype undergoing testing in the Alps.

More: VW’s Facelifted ID.4 Is Starting To Look A Lot Like A Tiguan

Disguised to look like the outgoing model, the electric Tiguan has a fully enclosed grille that will likely echo the one found on the new ID. Polo and facelifted ID.3 Neo . Bigger changes occur further below as the model has a new bumper, which channels air into revamped side curtains.

var adpushup = window.adpushup = window.adpushup || {que:[]};

adpushup.que.push(function() {

if (adpushup.config.platform !== "DESKTOP"){

adpushup.triggerAd("89c65a14-4682-4807-a4bb-5436c5361bee");

} else{

adpushup.triggerAd("f25825bd-1260-4574-84ce-2b26e315fad8");

}

});

The lower intake is new as well and it trades a honeycomb mesh pattern for horizontal bars. The crossover also sports a new center-mounted sensor pod.

Baldauf

Moving down the sides, the profile looks pretty familiar at first glance. However, a closer inspection reveals a surprising number of changes for a facelift.

In particular, there are new doors and a flatter shoulder line. Designers also added more conventional door handles as well as streamlined lower cladding. We can even see new third quarter glass and a reimagined rear pillar.

The extensive changes continue at the rear, where there’s an all-new liftgate, bumper, and diffuser. They’re joined by new taillights, which are hidden behind a photorealistic graphic.

Interior changes should be pretty significant as well and we can likely expect a new dashboard, upgraded displays, and revamped switchgear. We also wouldn’t be surprised to find a new steering wheel and higher quality materials.

var adpushup = window.adpushup = window.adpushup || {que:[]};

adpushup.que.push(function() {

if (adpushup.config.platform !== "DESKTOP"){

adpushup.triggerAd("bb7964e9-07de-4b06-a83e-ead35079d53c");

} else {

adpushup.triggerAd("9b1169d9-7a89-4971-a77f-1397f7588751");

}

});

The upgrades won’t end there as the ID. Tiguan will ride on the new MEB+ platform . We can also likely expect 58- and 79-kWh battery packs and a charging capacity of around 183 kW. They should be joined by upgraded motors offering improved performance and increased efficiency.

Of course, we’ll get a better idea of what to expect either late this year or early in 2027.

var adpushup = window.adpushup = window.adpushup || {que:[]};

adpushup.que.push(function() {

if (adpushup.config.platform !== "DESKTOP"){

adpushup.triggerAd("bb7964e9-07de-4b06-a83e-ead35079d53c");

} else {

adpushup.triggerAd("9b1169d9-7a89-4971-a77f-1397f7588751");

}

});

Baldauf

Full text

SAN FRANCISCO — AI giant Anthropic has already become a dominant player in technology and a household name for everyday users of artificial intelligence. Can it make drugs too?

On Tuesday, the company announced it is going to try. Though it’s unclear whether Anthropic intends to bring drug candidates to commercialization, multiple executives emphasized how important it is for Anthropic to get hands-on experience trying to use its own products to solve real scientific problems.

Eric Kauderer-Abrams, the company’s head of life sciences, said Anthropic has been asking itself what it should be doing besides training models and building products. During an event here to launch the company’s newest application, Claude Science , he said Anthropic had come up with one answer.

Continue to STAT+ to read the full story…

Full text

In a TikTok video that has racked up 1.4 million likes , men are handed scissors and a sheet of paper and asked to cut out a Christmas tree while describing their favorite memory with their partner. The men start telling a story, but most trail off as they concentrate on the cutting. In another clip, a woman calmly juggles four different chores while her husband struggles to put on socks and hold a conversation simultaneously.

These videos are designed to make you laugh. But they also reflect and reinforce the idea that women are better multitaskers than men. Studies putting this stereotype to the test have yielded very mixed results. Yet the idea that there is a gender difference in the ability to multitask remains lodged in the public mind. Researchers think there are various reasons why the stereotype persists.

Multitasking under the microscope

Science is clear on one thing: multitasking is incredibly difficult. The human brain has evolved to single-task. “We can do things at the same time when they are highly automatized, but other than that, humans in general struggle,” said Marco Hirnstein , professor in cognitive neuroscience at the University of Bergen, Norway. (I was relieved to hear that as I struggled to listen, take notes, and think of my next question all at once during our Zoom call.)

But when it comes to whether women are better multitaskers than men, scientists don’t seem to agree. Early studies seemed to back the stereotype. In 2013 , researchers in the U.K. found that everyone slowed down when multitasking — but men slowed down significantly more. Two years later, a Russian team reported that men used more parts of their brains to multitask and still took longer, suggesting that multitasking requires more effort in men.

Then the picture got complicated. A 2017 study concluded that men and women multitask equally well, even if their brains may go about it differently. In 2018 , when volunteers were given a realistic multitasking challenge — preparing a room for a meeting while fielding distractions — there was no difference between men and women on every measure the researchers tracked: accuracy, time, distance walked, remembering to do the right thing at the right moment, and dealing with distractions. A 2019 study found that multitasking led to a similar loss of speed and accuracy for both men and women.

Then a 2021 study flipped the script entirely, finding that men handled multitasking better than women, which the researchers chalked up to faster mental processing speed in men.

“The findings are incredibly inconsistent,” said Hirnstein . “I would say it’s probably even more inconsistent than you normally get in psychological research.”

Making sense of the research

Tilo Strobach , a professor in general psychology at MSH Medical School Hamburg, Germany, suspects that the mixed findings are due to differences in study design. “My feeling is that this mix is because of the different types of multitasking situations that were used across the studies,” he said.

Multitasking means doing two or more things simultaneously or in rapid succession. So talking while driving counts, but so does cooking, as it involves a quick sequence of steps. Different studies use different multitasking setups, with some task combinations overlapping in time and others being more sequential, explained Strobach. This variation in setups, he argues, helps explain why the findings are so mixed.

Gender differences in specific tasks also play a role. Hirnstein’s own research has shown that women tend to be better at finding and remembering words than men . Men are better at imagining what a complex 3D cube would look like if you were to rotate it . If you choose tasks for a multitasking experiment that already favor men or women, “then you’re probably just going to find the same difference that you observe in the single task,” Hirnstein said.

A good example of this is a 2013 study that found men were better at multitasking than women. The researchers traced this difference in multitasking outcomes to men’s superior spatial ability. Interestingly, when women were in the menstrual phase of their cycle — a time when women’s spatial skills are at their highest — the multitasking gap disappeared.

A self-fulfilling prophecy or a kernel of truth?

Strobach is certain that “there’s no difference between men and women at all” when it comes to multitasking ability.

But the stereotype can become a self-fulfilling prophecy, he said. If men expect to do worse than women, that expectation can sap their motivation and effort, increase their anxiety, and drag down their performance — making the stereotype look true even when no real difference in multitasking ability exists.

Hirnstein takes a slightly different view. He agrees that stereotypes affect performance, but believes that performance informs stereotypes too. People watch how others perform and generalize from that. The resulting stereotype is exaggerated and overly generalized, he said, “but there is this kernel of truth.”

Currently, “we don’t have proof that there is a sex difference in multitasking because the data is not consistent enough,” Hirnstein said. But he suspects there is one: “If you measure it right, you probably will find something.”

A specific behavior might drive the stereotype

André and Diana Szameitat, a husband-and-wife research team at Brunel University of London, have offered a different explanation for why the stereotype persists. André Szameitat is deputy director of the university’s Centre for Cognitive and Clinical Neuroscience; Diana Szameitat is an honorary lecturer in its psychology department.

In a study published earlier this year in Psychological Research , the Szameitats asked 41 men and 37 women to juggle five tasks at once in a setup that closely resembled real life. At one table, participants followed a recipe with pretend ingredients. When a kitchen timer rang, they broke off and dashed to a second table for two quick tasks — a phone-number search and a visual search — then returned to cooking. All the while, they watched a slideshow of words, jotting down any on a red background, a task that mimics the act of keeping an eye on a toddler. And every 20 seconds, a speaker played a recorded question they had to answer aloud, such as: “Would you rather lose all your money and valuables, or every photo you’ve ever taken, and why?”

Across the cooking, searching, and monitoring tasks, men and women performed equally well. However, a difference emerged in the talking task. Men failed to answer the questions more than twice as often as women.

When a group of neutral observers, unaware of what the researchers were testing, watched short films of the participants at work, they rated the female multitaskers as being more in control of the task, performing better, and being less stressed than the male multitaskers. Their judgment was highly influenced by the participants’ performance on the conversation task.

André Szameitat said the study showed two things. First, men and women differ in how well they multitask when one of the tasks is talking. Second, other people see this difference and use it to judge how well someone multitasks. “We suggest that this pattern can explain why the stereotype that women are better at multitasking than men may have developed,” he said. But he warned that the results only show a correlation for now, and more work is needed to prove that men’s reticence while multitasking is what drives the stereotype.

For now, it remains unclear whether men strategically drop the conversation to focus, or because they are unable to keep talking while doing something mentally demanding. “It could be a conscious decision,” Hirnstein said. “It could also be that they had so much going on at the same time that something had to give, and the first thing to go was this verbal task.”

If a male partner starts ignoring you mid-conversation while he’s busy with something else (just like the men in those TikTok videos), “one shouldn’t get angry,” Szameitat said. The study suggests this is a common pattern, not rudeness.

But there are settings where a man’s reduced talkativeness could be a serious issue. “There are highly critical situations where communication is essential, for example between pilot and co-pilot or control tower,” Szameitat said. Yet it is not yet known whether the findings would carry over into a high-stress event such as an aviation emergency, or whether the difference might disappear under that kind of pressure. This is a question for future research, Szameitat said. “Because these are early findings, it is important not to overstretch such implications and avoid headlines like ‘Men can’t fly planes’ or alike.”

This article The “women are better multitaskers” stereotype is messier than you think is featured on Big Think .

Full text

Stduy claims AI adoption outpaces governance, creating long-term code maintainability challenges

Saved developer time is now being spent on reviewing, validating and governing AI

Accountability and trust are now more important than speed and productivity

Even though 91% of organizations have two or more AI coding tools in active use, four-fifths (79%) of the more than 1,500 developers surveyed by GitLab believe software delivery hasn’t accelerated at the same pace as developer productivity, implying challenges along the workflow could be diminishing returns.

Around three in four believe developers are writing code faster (78%) and producing higher-quality code (73%), but a new report from the coding platform believes there’s much more to AI than speed alone.

GitLab describes this as an ‘AI paradox’, where developer time is being taken up reviewing, validating and governing AI despite its promised productivity impacts.

Is AI just shifting problems downstream?

In fact, 85% agree that the biggest constraint now is code review and validation, rather than code creation, proving that problems have simply been moved downstream instead of removed entirely.

But now that AI is deeply engrained within developer workflows, two in five (43%) now struggle to distinguish AI-generated code from human-written code, making it difficult to maintain security and quality in the long term. Three in four (73%) say they’re concerned about the long-term maintainability of AI code.

And it’s this visibility that’s causing one of the biggest headaches for coders, with a third (34%) now unable to determine whether AI-generated code played a role in an incident.

“The events of the past few months, including supply chain attacks, reliability issues, and regulators tightening expectations around AI traceability and provenance are making clear that speed without control is a liability, not an advantage,” Chief Product and Marketing Officer Manav Khurana commented.

Most companies (92%) now experience some form of governance challenges when it comes to vibe coding, and four in five admit they’ve adopted AI coding tools faster than they’ve implemented governance policies.

Trust to become a key differentiator – not speed

While coding has become a major use case for generative AI since its mainstream introduction in late 2022, policies are still falling behind. But developers know this, and 91% now plan to invest in governance over the next year, with 98% allocating budget specifically for this.

Looking ahead, developers are preparing to enter a new era of AI, where generating code is no longer a priority. Instead, it’s all about governance.

“Organisations that will ship trusted software faster are the ones building the foundations of accountability with context, traceability, and governance baked into the platform, not just bolted on after the fact,” Khurana added.

But reading between the lines, it’s been clear that developers haven’t been entirely comfortable with AI for some while. Stack Overflow’s 2025 developer survey found that many are turning their backs on the tech due to privacy, pricing and quality reasons. Nearly 46% distrust AI to some degree, compared with just 33% who trust it.

Developers have now reached a point where AI has very clear use cases, but GitLab’s report clearly shows that traceability, accountability and trust will become competitive advantages in the future.

Full text

After a long hiatus from Civilization 7 following my original review – after which I wasn't exactly in love with it – I decided to give it another go with the big 1.4 Test of Time update, and more specifically the 1.4.1 patch that came out in late June 2026. Firaxis has made pretty big changes across the board, including some much-needed quality-of-life improvements here. But I don't know if I agree with the overall philosophy of making it more like older, existing Civ games.

In May, Take-Two' Strauss Zelnick told Game File that Civilization 7 had been a "bridge too far" for existing fans of the series. That's definitely the guiding star the Test of Time update was based on. And man, from my perspective, that was completely the wrong message to take from the negative feedback to Civ 7's launch. There's a quote I keep coming back to from No Man's Sky developer Sean Murray that goes, "Players are normally almost always right about problems … [but] they're almost never right about the solutions." I think Firaxis has put too much faith in the solutions certain vocal portions of the player base were calling for.

What I said about Civilization 7

With Civilization 7, the desire to streamline and simplify this legendary 4X series feels like it has gone a bit too far, particularly when it comes to the interface. It’s frustratingly inadequate at providing me with the information I need to play well, or even understand what's going on sometimes. Even so, it does have improved warfare and diplomacy, a bit of added narrative flair, as well as mostly gorgeous graphics and sound, so it can still give me reasons to keep clicking one more turn late into the night. There’s good reason to believe that with Firaxis’s usual pace of patches and expansions it can refine its new ideas and become everything it ought to be, and while what’s here right now is at least still a fun time, it’s also a bigger step backward for the series than we’re accustomed to when a new age dawns. - Leana Hafer, February 3, 2025

Score: 7

Read the full Civilization 7 review .

I'm a fan of civ switching. It's very contentious, and I get that. And it hasn't gone away or anything. You can still switch civilizations every age if you want to. You can set the AI to change civilizations every age. I don't even hate the way "timeless" civs have been implemented, necessarily – which let you play the same civ through all three ages like in previous games. There are unique policy cards for each one even when they're not in their native "apex" age. And I especially love the Syncretism feature that lets you borrow unique units or improvements from a civ you're aware of that’s in its apex age.

I played one campaign as Iceland through all three ages, and in the modern age I picked up American Marines as a unique unit through Syncretism, which felt very appropriate. You know, we used to be vikings, so of course we'd be good at amphibious warfare. I don't hate that it's an option. It's more the principle of the thing. I don't feel like Civ 7 puts its best foot forward when it backs off of some of its most interesting ideas.

Bittersweet Victory

Where this really turns out to be a bummer is in the victory conditions, with each one having been converted into a really simplistic race for points that may or may not include some vestigial nods to the age-specific objectives from 1.0. A lot of those objectives were a total mess. I'm not denying that at all. But rather than getting refined, they've been replaced with something far less interesting.

Tourism is back, but it doesn't work anything like it did in Civ 6. It's just the name for cultural victory points, and it hasn't done anything to break up the wonder spam meta. There's also Domination for military, GDP for economic, and Innovation for science, and they're functionally almost identical. The science victory has a small, interesting wrinkle that you need to build and defend a launch pad.

But all of the victories now work by requiring you to get your score in one of the four categories to a certain threshold above the civ in second place, which starts a five turn countdown before ending the game. 1.0 mechanics like factories and treasure fleets now simply give you GDP points, which also looks at things like how much gold your cities are producing and how many trade routes you have. Domination mostly comes from owning conquered cities. Innovation comes from special projects and the tech tree.

The victory threshold shrinks over time from 5x in Antiquity to, at the lowest I've seen, 1.5x about 60% of the way through the Modern Age. This means you can end the campaign quite early if no one else is really competing with you for the same victory objective. And while that can be a welcome way to wrap things up instead of playing out 50 more turns of a foregone conclusion, it also feels like an admission that they just couldn't make the late game fun, so they might as well give you a way to skip it.

The 1.4 patch notes talk about the fact that you can now focus on one victory condition from the very beginning like it's a huge improvement, but I actually liked the ability to change up my strategy between ages, too. And there's no longer any reward for doing that. My Icelandic Empire was the terror of the seas during the Exploration Age, as they should be, and I just ended up continuing my momentum into a Domination victory in the Modern Age even though it might have been more interesting to settle down and go for Tourism with all my bonuses to natural wonders. It was too late to change my mind, really.

Sneaky-pedia

The quality-of-life changes are significant, at least. The interface and tooltips have been greatly expanded with ways to quickly access more information. You can pin tooltips using… the K key? Okay. Weird choice, but you can rebind it at least. The Civlopedia is still missing plenty of features I'd want, like clickable hyperlinks that take you to a relevant article. But it's a lot easier to locate relevant details or dig deeper for more information on how a specific figure is calculated.

Game setup options have been expanded as well, and are a bit better explained. They're still not quite as robust as older Civ games. But certain map types come with additional variables you can tweak like sea level. Archipelago actually looks like an archipelago, though I still think it's a bit too land dense on the high sea level setting. There's a bunch of custom difficulty modifiers as well, if you don't mind the AI getting free money but don't want their units to get arbitrary combat bonuses, which is nice. You can even set whether they stick with their starting civilization or pick a new one every age.

The Test of Time update didn't suddenly make this my go-to Civ.

So where am I at with Civ 7 after all these updates? Does it finally have the juice? Well, I started a new campaign while I was writing this that I plan to keep playing just for fun, so like I said in my initial review, it's not lacking in that “just one more turn” quality the series is known for. It's a huge relief to have an interface that doesn't hide important details from me. But here's the thing: I would still rather be playing Civ 5 or Civ 6. Civ 7 still isn’t able to compete with their years of patches and expansions, especially since it has backtracked on some of its more daring ideas, which makes it feel even more like a stripped-down Civ 6.

The Test of Time update didn't suddenly make this my go-to Civ. I think it's going to take at least a big expansion or two to find out if it ever crosses that line. And while I disagree with the design philosophy that has guided a lot of its changes since launch – victory conditions in particular feel like a step back – there are at least meaningful improvements I can point to as well.