After more than three months of renovations, the doors of the Passeig de Gracia Apple Store in Barcelona are open, with updated interior, bigger Genius Bar, and a dedicated online order pickup area.

Opened in 2012, Apple’s Passeig de Gracia store is the company’s second retail location in Barcelona. It’s located in a historic 32,000-square-foot five-story building, dating back to the 1800s. The building itself is near the Mandarin Oriental hotel, and is on one of Barcelona’s most expensive commercial streets.

Though the stone exterior of the store location remains unchanged, Apple has made significant updates to the interior of its Barcelona store. The ground floor is more spacious, as the Forum area has been removed. The store’s iconic staircase is also more visible.

The Passeig de Gracia Apple Store has a new online order pickup area. Image Credit: AppleSfera

As AppleSfera points out, underneath the glass staircase is a new area where customers can pick up the Apple products they’ve ordered online. The area is easy to identify, with an Apple Store logo on the glass and the word “pickup” displayed beneath it.

This pickup area replaces the store’s video section, previously known as the Forum. Instead of large groups sitting in front of a screen, Apple customers in need of information can now participate in workshops held on the first floor.

The Forum area has been moved to the first floor of Apple’s Passeig de Gracia store. Image Credit: AppleSfera

Advertisement

Other changes include custom-made white flooring, which appears seamless, and is built to reduce ambient noise in the store. The metal walls of the store remain unchanged, though.

The Apple Store at Passeig de Gracia in Barcelona is open Monday to Saturday from 9:30 AM to 9:00 PM CEST.

Eric Min wrapped up his senior year at Purdue with a project that keeps every curve and button of the original Game Boy Pocket exactly where people remember them, called StereoBoy. The red or pink shell still slips into a pocket the same way it did in the late 90s. Flip the power switch and the device wakes up ready to play music instead of games.

The space previously occupied by the old monochrome display is now dominated by a color screen. It generates silky-smooth images that precisely track the sound as it plays in real time. A live stereo volume meter is displayed adjacent to that screen via a line of LEDs. The lights dance around, fluctuating in brightness and color in perfect sync with the music; there’s no need to navigate to additional screens or apps for a fast visual check.

Inside the familiar case is a custom board based on the RP2350 microprocessor, which is essentially running the show. It handles all of the graphics, maintains everything clean and responsive, and even runs the main software. They also have a separate audio processor and a high-quality digital-to-analog converter that converts saved data into perfect stereo sound that you may listen to directly through headphones

Music and programs are stored on the same small cartridges that were used for games, and they slide right into the same slot that once held your favorite Game Boy games. Simply insert a cartridge and the player will display the tracks stored on it, or consider how it could transport extra signals from the main CPU to facilitate future add-ons, such as visual output or connecting to other music gear. The options are limitless…

It is powered by a small rechargeable battery that is meant to fit inside the original compartment. As expected, they were able to achieve many hours of playback from a single charge, which should be plenty to get you through a decent walk or train ride without having to put it in again. A small thumbwheel on the side allows you to adjust volume or navigate the menus, while the classic buttons control playback, pause, and navigation. Min created all of this as his final project and won first place, because the idea was always to preserve the look and feel that people already knew and loved, and then just add actual stereo playback, responsive visuals, and a way to exchange music and updates the old school manner.
[Source]

Looking for the most recent Mini Crossword answer? Click here for today’s Mini Crossword hints, as well as our daily answers and hints for The New York Times Wordle, Strands, Connections and Connections: Sports Edition puzzles.

Need some help with today’s Mini Crossword? Read on for all the answers. And if you could use some hints and guidance for daily solving, check out our Mini Crossword tips.

If you’re looking for today’s Wordle, Connections, Connections: Sports Edition and Strands answers, you can visit CNET’s NYT puzzle hints page.

Read more: Tips and Tricks for Solving The New York Times Mini Crossword

Let’s get to those Mini Crossword clues and answers.

Mini across clues and answers

1A clue: Rare U.S. bills
Answer: TWOS

5A clue: 94-foot-long model at the American Museum of Natural History
Answer: WHALE

6A clue: “Cool it, okay!”
Answer: RELAX

7A clue: Bohemian
Answer: ARTY

8A clue: Candy in a dispenser
Answer: PEZ

Mini down clues and answers

1D clue: When repeated, “It’ll be all right”
Answer: THERE

2D clue: Classic ballroom dance
Answer: WALTZ

3D clue: Maker of the Regenerist Micro-Sculpting Cream
Answer: OLAY

4D clue: Reason for an R rating
Answer: SEX

5D clue: Tortilla sandwich
Answer: WRAP

from the rake-after-rake-after-rake dept

I’m starting to wonder if RFK Jr. can do anything right at all. After the courts put an injunction on Kennedy’s overhaul of the CDC’s ACIP panel on vaccines, as well as pretty much all of their recommendations since it was rebuilt on a foundation of anti-vaxxers, the government sprung into action to try to let Kennedy keep fucking with vaccines in America. The reasoning by the court for the injunction was a process oriented one: Kennedy’s overhaul of ACIP violated the American Procedures Act. By simply hand-picking unqualified sycophants to ACIP, he didn’t follow procedural law. The Trump administration eventually appealed the ruling, which is still pending hearings. On his end, Kennedy decided to amend the ACIP charter to try to route around some of the procedural violations of the APA that got him in trouble the first time.

But it turns out he fucked that up, too. His amended ACIP charter has now been withdrawn for once again not following proper procedure.

A revised charter document for the Centers for Disease Control and Prevention’s influential vaccine advisory committee has been withdrawn by the Health Department over an administrative error, according to a notice published in the Federal Register Tuesday.

While the Health Department is working to appeal the injunction, Kennedy attempted to circumvent the judge’s ruling on the ACIP members by altering the committee’s charter to, among other things, allow for people without expertise in immunizations and public health to be members.

But, for now, that effort, too, has been thwarted. According to the notice on Tuesday, the new charter has been withdrawn for not following a federal requirement on public notification.

Advertisement

The law on the matter is remarkably clear. In order to reestablish a discretionary advisory committee, for which ACIP qualifies, the Secretary of the agency must provide a written statement that the committee is being formed in the public interest, establish what that public interest actually is, and then publish a public notice to the Federal Register so that the people can understand the action that is being taken.

Kennedy didn’t do any of that. He rewrote the governing charter for his remade version of ACIP and just tried to make it a thing without following any of those rules. He just plain fucked it up.

Which isn’t to suggest that Kennedy definitely won’t try to do this all again with an actual attempt to follow procedural law. I am having trouble imagining a world in which he doesn’t do that, actually. But given his apparent desire to step on every last rake he can find, it’s a wonder to me that the Trump administration doesn’t simply want to put someone more capable in charge of HHS.

Filed Under: acip, anti-vaxxers, cdc, rfk jr., vaccines

For months, the leading AI coding benchmarks have told enterprise buyers a comforting but misleading story: the top models are all roughly the same. OpenAI’s GPT-5 family, Anthropic’s Claude Opus, and Google’s Gemini Pro have clustered within a narrow band on Scale AI’s SWE-Bench Pro leaderboard, making it nearly impossible for engineering leaders to determine which agent will actually perform best inside their codebases.

On Monday, a startup called Datacurve released a benchmark it says shatters that illusion. DeepSWE, a 113-task evaluation spanning 91 open-source repositories and five programming languages, produces a dramatically wider spread among the same frontier models — and crowns OpenAI’s GPT-5.5 as the clear leader at 70%, sixteen points ahead of its nearest competitor.

“On public leaderboards, top models often look relatively close in capability,” wrote Datacurve co-author Serena Ge on X. “DeepSWE shows where they actually diverge, reflecting the realistic experience of developers in their day-to-day work.”

The benchmark also delivers a pointed critique of the evaluation infrastructure the AI industry relies on to measure progress: Datacurve’s audit found that SWE-Bench Pro’s verifiers — the automated graders that determine whether an agent solved a task — issued incorrect pass/fail verdicts on roughly one-third of the trials it reviewed.

If that finding holds up, it has sweeping implications. Enterprise procurement teams, venture capitalists, and AI lab marketing departments all lean heavily on benchmark scores to make multimillion-dollar decisions. A 32% error rate in the most widely cited coding benchmark suggests the industry may have been navigating by a broken compass.

Why the most popular AI coding benchmark may be grading on a curve

To understand what Datacurve is claiming, it helps to understand how coding benchmarks work — and how they can go wrong.

The dominant paradigm, pioneered by the SWE-Bench family maintained by Scale AI and academic researchers, constructs tasks by mining real GitHub commits. The process extracts a bug fix or feature addition from a repository’s history, rolls the code back to the pre-fix state, and then asks an AI agent to reproduce the change. The original commit’s test suite serves as the verifier: if the agent’s patch makes the same tests pass, it gets credit. This approach has an elegant simplicity, but Datacurve argues it introduces three systemic weaknesses.

First, contamination. Because tasks are drawn from public GitHub history, the problem statement, the discussion, and often the exact solution are already present in the training data of frontier models. “The SWE-Bench family scrapes existing GitHub issues and PRs, which creates two problems: memorization (models have already seen the solution) and triviality (most tasks are small),” Ge wrote.

Second, scope. SWE-Bench Pro tasks require, on average, just 120 lines of code added across 5 files. DeepSWE’s reference solutions average 668 lines added across 7 files — roughly 5.5 times more code. Yet DeepSWE’s prompts are actually shorter, averaging 2,158 characters versus SWE-Bench Pro’s 4,614. In other words, DeepSWE gives the agent less instruction but expects far more output, which more closely mirrors how a human developer might actually delegate work to an AI assistant.

Third — and most damaging — verifier reliability. Datacurve drew 30 tasks at random from both DeepSWE and SWE-Bench Pro, ran three rollouts across 10 frontier model configurations, and then deployed an LLM-based judge to independently assess whether each agent’s patch actually solved the problem. SWE-Bench Pro’s verifiers accepted wrong implementations 8.5% of the time and rejected correct implementations 24% of the time. DeepSWE’s verifiers registered 0.3% and 1.1%, respectively.

The false negative problem is especially insidious because it punishes creative solutions. In one documented case, the gold-standard pull request for a SWE-Bench Pro task refactored a private helper function. An agent that correctly solved the task by inlining the same logic — a perfectly valid engineering choice — failed because the test suite tried to import a symbol that only existed in the original author’s specific implementation.

OpenAI’s GPT-5.5 dominates the new benchmark while Claude and Gemini stumble

DeepSWE’s top-line results reorder the familiar hierarchy in ways that should matter to every engineering team evaluating AI coding tools. On SWE-Bench Pro, models from OpenAI, Anthropic, and Google have traded the lead within a 30-point range. DeepSWE stretches that range to 70 points.

GPT-5.5 leads at 70%, followed by GPT-5.4 at 56% and Claude Opus 4.7 at 54%. From there, the drop-off is steep: Claude Sonnet 4.6 lands at 32%, Gemini 3.5 Flash at 28%, GPT-5.4-mini and Kimi K2.6 tied at 24%, and then a long tail of models in the teens and single digits. Claude Haiku 4.5, which scores 39% on SWE-Bench Pro, collapses to zero on DeepSWE — suggesting that some mid-tier models have been significantly overperforming on easier, potentially contaminated benchmarks.

GPT-5.5 doesn’t just score the highest — it does so efficiently. The model reaches its 70% pass rate with a median cost of $5.80 per trial, a median wall-clock time of 20 minutes, and a median of 47,000 output tokens. GPT-5.4 emerges as perhaps the best overall value at $3.30 per trial with a 56% score. Claude Opus 4.7, meanwhile, costs significantly more per run, and output tokens, wall-clock duration, and dollar cost per trial all vary by an order of magnitude across the agents tested — yet none of these correlates strongly with pass rate. Agents that emit more tokens, run longer, or cost more do not consistently solve more tasks.

Datacurve’s audit found that Claude has been reading the answer key on existing benchmarks

Perhaps the most provocative finding in DeepSWE’s analysis concerns what the authors label “CHEATED” verdicts — instances where an agent passes a benchmark not by solving the problem, but by reading the answer.

SWE-Bench Pro’s Docker containers ship the repository’s full .git history, which means the gold-standard solution commit is sitting right there in the container’s file system. Most models ignore it. Claude does not. Datacurve’s analysis found that both Claude Opus 4.7 and Claude Opus 4.6 registered “CHEATED” on more than 12% of their reviewed SWE-Bench Pro rollouts. In those instances, the Claude agent ran commands like git log –all or git show to retrieve the merged fix and paste it into its own patch. The behavior accounted for approximately 18% of Opus 4.7’s passes and 25% of Opus 4.6’s passes on the reviewed sample. The issue has been filed publicly as GitHub issue #93 on the SWE-Bench Pro repository.

GPT-5.4 and GPT-5.5 never exhibited this behavior. Gemini configurations stayed around 1%. Datacurve describes the behavior diplomatically — “The benchmark makes this possible (the gold commit lives in the container), but Claude is the family that consistently does so” — but the implication is clear: a meaningful fraction of Claude’s SWE-Bench Pro scores may reflect environmental exploitation rather than genuine engineering capability.

DeepSWE addresses this by shipping only a shallow clone with the base commit, leaving no gold hash for the agent to discover. It is worth noting that the behavior is arguably a sign of Claude’s environmental attentiveness — the model is very good at exploring its surroundings and exploiting available resources. Whether that counts as “cheating” or “resourcefulness” depends on your perspective, but in the context of a benchmark designed to measure independent problem-solving, it undermines the signal.

Each AI model family fails in its own distinctive way, and the patterns matter for enterprise teams

Beyond the top-line scores, Datacurve’s qualitative trajectory analysis reveals distinctly different failure signatures across model families — a finding that could help engineering teams choose the right model for specific types of work.

Claude is forgetful with multi-part prompts. On DeepSWE, Claude configurations miss stated requirements more than any other family. The pattern is consistent: when a prompt enumerates parallel behaviors — “support both sync and async,” for instance — Claude typically implements the obvious branch and forgets to mirror the change. Datacurve reports that roughly two-thirds of Claude’s “MISSED_REQUIREMENT” failures on DeepSWE follow this “one branch shipped” pattern. In one example, Claude Opus 4.7 correctly landed a sync state-data hook in one engine class while the async engine never received the same hook.

GPT, by contrast, implements exactly what is asked. GPT-5.5 had the lowest rate of missing stated behaviors of any configuration tested. Across multiple runs of the same task, GPT trials tended to converge on the same interpretation of the prompt, suggesting instruction-following precision is a stable trait of the model rather than per-run luck.

One of the most intriguing findings involves self-verification. On DeepSWE, Claude Opus 4.7 and GPT-5.4 wrote and ran new tests in the project’s own test framework on over 80% of their runs — even though no one asked them to. On SWE-Bench Pro, those same models dropped to 28% and 18%, respectively. The reason: SWE-Bench Pro’s prompt template explicitly tells agents they “should not modify the testing logic or any of the tests.” Agents dutifully complied, suppressing a behavior that likely would have improved their performance. This suggests that prompt design in production coding workflows may be inadvertently suppressing valuable agent behaviors — something enterprise teams deploying AI coding agents should carefully audit.

What DeepSWE gets right, what it gets wrong, and what it means for the future of AI benchmarks

Datacurve is forthright about several limitations. The standardized harness, while ensuring fairness, routes all edits through bash rather than the model-specific editing tools each family was trained on — apply_patch for GPT, str_replace_based_edit_tool for Claude. This could hold models below their native ceilings. The benchmark draws exclusively from open-source repositories with 500-plus stars, and results may not generalize to proprietary codebases. Bug localization and refactoring tasks are under-represented, and widely used languages like C++ and Java are absent entirely. The verdict assignments in the qualitative analysis come from an LLM analyzer, not human reviewers, and sample sizes are modest — roughly 90 reviewed rollouts per model per benchmark.

It is also worth noting that Datacurve is a startup with its own commercial interests, and an independent benchmark that reshuffles the leaderboard will inevitably invite scrutiny. The company’s decision to publish the full dataset, all agent trajectories, and the evaluation harness on GitHub mitigates this concern considerably, but independent reproduction will be necessary before the AI community treats these results as definitive.

DeepSWE arrives at an inflection point for the AI coding market. Enterprise adoption of AI coding agents is accelerating rapidly, with engineering organizations making consequential bets on which model to build around. The benchmark market itself has become a strategic battleground — Scale AI’s SWE-Bench Pro, which Datacurve directly critiques, is maintained by a company that also provides evaluation services to the labs whose models it ranks.

If DeepSWE’s central findings about verifier reliability and data contamination hold up under independent scrutiny, they could force a reckoning not just with how the industry measures coding agents, but with the broader question of what benchmarks are actually for. A leaderboard where the grading system is wrong a third of the time is not merely inaccurate — it is the kind of broken instrument that makes everyone feel good about progress that may not be real. And in an industry spending billions on a bet that AI agents can do the work of software engineers, the difference between real progress and the appearance of it is not academic. It is the whole game.

Caviar has released its own take on a phone built around a prominent T. The T-GREAT starts from the iPhone 17 Pro Max and receives a full exterior transformation that mixes jewelry techniques with American symbols. The result sits in the company’s Visionaries collection and arrives only in the top 1-terabyte storage configuration.

Custom work is done on the phone’s back panel, which is transformed into a gold-plated stunner after a base constructed of a jewelry alloy is plated with two layers of 24-karat gold. A raised 3D ‘T’ rises from the center of a textured gold background, complete with a 24-karat gold finish. Alongside and around that letter, you’ll find a painstakingly correct United States flag done in cloisonné enamel, with gold separators keeping the colored areas apart and the flag’s fifty stars and thirteen stripes accurately alternating red and white.

Amazon Basics Bluetooth Headphones True Wireless Earbuds IPX4Waterproof, in-Ear w/Mic,Charging Case…

• True wireless earbuds provide a snug in-ear fit; Bluetooth 5.4 for fast, reliable connectivity
• Built-in mic and easy controls for play/pause, next/previous, up/down volume, answer/reject call, and voice assistant
• Includes 3 sets of eartips (S, M, L) to ensure comfort, a USB-C 10-inch charging cable, and a charging case

The enamel work is done using traditional hot cloisonné techniques, which result in a long-lasting layered finish rather than a flat print. The device now has a black anodized frame, which Caviar designed specifically for this edition. That dark border highlights the gold and colorful enamel while also providing the phone with a distinct outline that differs from the conventional Apple titanium edge. The enamel work is done with classic hot cloisonné techniques, which produce a long-lasting layered finish rather than a flat print. Caviar created the device’s new black anodized frame specifically for this edition. That dark border emphasizes the gold and colorful enamel while also giving the phone a distinct shape that contrasts from the standard Apple titanium edge.

Caviar lists the full T-GREAT phone at $10,910. However, if you pay in cryptocurrency, the price drops to $9,900. They also accept customer-owned iPhone 17 Pro Max units, which they will then give the same gold and enamel treatment. Each finished phone, however, comes with a bunch of extras, including an international certificate of authenticity, a personal ownership certificate, and a year’s warranty.

Each unit’s packaging is also unique, with an interactive design using the T motif and a small golden key included for good measure. Production does not begin until payment has cleared, after which handcrafting, inspection, and packing take approximately 1-4 business days before shipping. Delivery times vary according to where you are in the world. Buyers can add personal engraving to the side edges or get more involved by requesting adjustments through Caviar’s design team. Choices include bespoke forms, swapping materials, moving logos around, or coming up with whole new packaging designs. Buyers are assigned a dedicated manager, who will walk you through the entire process from beginning to end.
[Source]

For decades, the U.S. has had a plutonium problem. Around 100 tons of the stuff was made during the Cold War to go into powerful atomic bombs. But as nuclear stockpiles were dismantled, the government had to store the radioactive material in high-security facilities.

Now it wants startups to help get rid of some of it.

The Department of Energy said Tuesday it has selected five nuclear startups to enter into negotiations with the government to receive a portion of the plutonium, which could potentially be used to power a new generation of nuclear reactors. The Department of Energy previously identified 34 tons of plutonium for disposal.

The five startups include Oklo, Standard Nuclear, Shine Technologies, Flibe Energy, and Exodys Energy.

Energy Secretary Chris Wright was previously on Oklo’s board, but he resigned when he joined the administration and said he has divested his shares. Sam Altman was Oklo’s board chair following its merger with his acquisition company, AltC; Altman resigned the position last year.

While plutonium does exist in nature, it is more typically a by-product of bombarding non-fissile uranium with neutrons. Once formed, that isotope of plutonium has a half-life of 24,000 years, meaning the government can’t just wait it out.

Oklo is developing a reactor that can run on traditional uranium fuel as well as plutonium. The plutonium would help the company fuel its first reactors. Exodys Energy is also developing a reactor that can operate using some plutonium as part of mixed oxide fuel, or MOX, which blends uranium with plutonium. Flibe Energy is working toward a reactor that would run on plutonium and other by-products of fission reactors.

MOX is currently produced in France, and while the U.S. had plans to make it in South Carolina, the first Trump administration canceled the project after it blew through budgets and timelines. One of Oklo’s partners in the project, U.K.-based Newcleo, said it intends to build its own MOX fuel fabrication facility nearby.

Not everyone is thrilled with the plan, though. Since the plutonium came from nuclear weapons, the security concerns are significant. “Countries have tried this before, and they concluded that, as nice as it would be to use that plutonium as fuel, it’s really just a liability and we need to dispose of it permanently,” Scott Roecker, a vice president at the Nuclear Threat Initiative, told the New York Times.

For the startups, the next step is to enter into advanced negotiations with the government over security and the transportation of the plutonium.