RSS Feeds

In this disorienting book, a poet argues we have been looking at nature – and ourselves – the wrong way

The eastern breakwater of Granton harbour is almost 1km long, and swimmers can follow its crooked finger, heading straight out into the Firth of Forth, while benefiting from its reassuring presence. One Thursday lunchtime, I thought: why not, let’s give it a go, and headed for its tip. I felt cold, but, as I progressed, I seemed to grow strangely warmer. I did not notice the complaints from my body: the creeping numbness which proceeded from my fingers to my elbow, nor the coldness of my pee. Only by the time I exited the water, club footed and claw handed, did it hit me that I was hypothermic.

The irony of this reckless exercise was that in my swim bag there was a book by Melanie Challenger called Alive: The Hidden Intelligence of the Living World. It’s an ambitious examination of the wisdom of bodies, a rhizomatic guide not just to our own, but also all the living forms around us: from bacteria to chimps. Her case is that we have ignored or misread the agency and purpose of bodies to our detriment; in a way it is a plea to listen to them. I wish I had read it before my ill-advised swim.

Authors: Ying Li, Arjun Rao, Shradha Sehgal

Introduction

Recommendations sit at the heart of the Netflix experience. Our current production models rely on thousands of hand‑crafted features over users, items, and interactions, along with specialized architectures for sequence modeling, feature interactions, and multi‑task objectives. This stack has evolved over many years to support diverse content types (movies, series, games, live, podcasts) and product surfaces, but its complexity makes it costly to onboard new use cases: adding a content type or surface can require significant feature engineering, architecture change, infrastructure work, and experimentation.

At the same time, large language models (LLMs) are changing how we think about recommendation, as shown by recent work such as PLUM, GLIDE, and OneRec-Think. Their broad world knowledge and strong language understanding make it possible to represent user histories and item metadata directly as text, capture rich relationships in a shared semantic space, and steer recommendations via natural‑language prompts. However, off‑the‑shelf LLMs are still far from production‑ready recommenders: they often over‑recommend globally popular content, hallucinate out‑of‑catalog items, ignore business constraints, and provide only limited personalization.

To address this, we built GenRec, an LLM‑backed recommendation ranker that post‑trains an internal foundation LLM on Netflix‑specific data and objectives. GenRec shows that an LLM‑based ranker can match or exceed a mature production system while relying on far fewer labeled examples and input signals.

At a high level, GenRec:

• Verbalizes user histories, item metadata, and context as text.
• Post‑trains a Netflix‑adapted foundation LLM for ranking.
• Adds a catalog‑aware scoring head over Netflix titles.
• Uses reward signals to align with long‑term member value and business goals.
• Runs in prefill‑only mode on Netflix’s LLM serving stack for cost efficiency.

In a large‑scale A/B test against a well‑tuned production ranker, GenRec achieves statistically significant improvements in both short‑term and long‑term online metrics, while using only a small fraction of the Phase‑2 labeled data and input signals. It reduces our reliance on hand‑engineered features and shifts the focus from feature engineering to context engineering. In this blog post, we will describe how GenRec works, how it performs, and why we believe it points toward a more LLM‑centric future for recommendation at Netflix.

Problem Setting

We focus on a full‑catalog ranking task (or top‑K ranking when a candidate set is provided).

Given a user 𝑢, their interaction history 𝐻, and the current context 𝜏 (device, surface, locale, time, etc.), GenRec scores each item and produces a personalized ranking that can directly power recommendations or serve as input for downstream personalization systems.

Formally, we map a request (u,τ,t,H) — user, context, time, and history — to a ranking 𝜋 over the catalog C, where π(i) is the position assigned to item i. We optimize π for expected long‑term member utility (a proxy for satisfaction and retention), not just short‑term engagements.

From Foundation LLM to Recommendation Ranker

GenRec follows a two‑phase training framework (Figure 2):

Phase 1 — Netflix-Adapted Foundation LLM.

We start from an open‑source LLM and adapt it on proprietary Netflix corpora, so it learns foundational capabilities such as

• Netflix content understanding
• Member behavior and preference patterns
• General language understanding and generation.

Phase 1 is updated relatively infrequently and serves as a shared, Netflix‑aware backbone for many applications.

Phase 2 — GenRec.

We then turn this foundation model into a high‑quality ranking model by post‑training on ranking‑specific data and objectives. Phase 2:

• Focuses on ranking quality and steering
• Incorporates multiple reward signals via reward‑weighted losses
• Is refreshed more frequently to track new content and evolving tastes
• Is explicitly optimized under serving cost constraints.

Training Data as Conversations

Netflix members generate hundreds of billions of interaction events spanning many surfaces (views, plays, durations, thumbs up/down, add to list, abandons, etc.). We convert this log data into single‑turn or multi‑turn “conversations” between a user and a recommender. Each turn contains:

• User message: verbalized context, profile, history, item metadata, and task (e.g., recommend what the user will watch or thumb next).
• Assistant message: the member’s actual engagement (e.g., which titles were played, for how long, what feedback they provided).

During Phase‑2 training, the LLM learns how assistant messages depend on user messages. This allows us to express rich recommendation signals as text, jointly supporting both the language-modeling (LM) and ranking objectives.

At inference time, we feed in the verbalized context and apply a catalog‑aware scoring head to rank items; we do not decode assistant messages. The conversational format is primarily used during training to support the LM objective and preserve strong language understanding over the verbalized text.

Verbalization and Context Engineering

Traditional recommenders operate on dense features and embeddings. GenRec takes a different approach: it verbalizes rich user histories and context as natural language, encoding raw interaction signals directly in the LLM’s semantic space. In doing so, it relies on the model to discover higher‑level patterns — such as item relationships and evolving user interests — rather than on manual feature engineering.

Naively verbalizing every interaction in a user’s history can quickly exceed the token budget and be too expensive at Netflix scale. The context window becomes our new “feature budget”, so we apply context engineering:

• Retain in full: high‑signal engagements (e.g., long plays, thumbs‑up) with richer details
• Omit: low‑signal events (e.g., very short plays or quick hovers)
• Summarize or compress: repetitive behaviors (e.g., binge‑watching )
• Elaborate selectively: important or cold‑start items (e.g., new releases)

Within a fixed token budget, we prioritize recent, high‑signal history and compress or drop older history. We also structure the prompt to maximize shared prefixes for better prefix caching. The goal is a compact, high‑information prompt that preserves ranking quality without prohibitive costs.

Objectives: Ranking, Language, and Rewards

The overall GenRec model is trained with a multi‑objective loss that combines a recommendation ranking objective, language modeling objectives, and alignment via reward‑weighted training.

1. Catalog‑Aware Ranking Objective

The primary task is a ranking objective that teaches the model to score items by engagement quality. We label positives using high‑value engagements (e.g., sufficiently long plays, strong explicit feedback), with thresholds and denoising logic, and train the model — via a cross‑entropy loss over the catalog or candidate set — to assign higher scores to these positives given a verbalized context.

2. Language Modeling Objective

We also retain a language modeling (LM) objective over the verbalized inputs and outputs. This helps preserve the model’s general language understanding, improves its ability to interpret rich natural‑language histories and item metadata, and keeps the door open for text‑generation use cases such as recommendation explanations.

3. Reward‑Weighted Loss for Alignment

Beyond raw ranking accuracy, GenRec must (1) respect business requirements — for example, balancing movies, series, games, live, and podcasts — and (2) optimize long‑term member satisfaction rather than just immediate clicks or plays.

Training only on raw interaction sequences can lead to undesirable behaviors, such as over‑favoring binge‑watching or over‑focusing on a single content type. To address this, we weight the ranking loss using signals from separate reward models. Each training example receives a scalar weight derived from two types of signals:

• Long‑term satisfaction proxies: estimate how much a short‑term engagement contributes to long‑term outcomes, such as return behavior, catalog exploration, or sustained engagement.
• Behavior rebalancing: adjust behaviors across content types and launch stages (for example, games vs. movies, new releases vs. evergreen titles) to better align with business goals.

The example’s ranking loss is then scaled by this weight: high‑value engagements receive larger weights, and low‑value ones are down‑weighted. This reward‑weighted approach is simpler and more cost-efficient than full reinforcement learning, yet provides effective alignment in practice. We have seen additional gains from RL‑style methods (e.g., GRPO), but leave them to future work due to their higher cost.

Model Architecture and Serving

Backbone and Scoring Head

GenRec’s architecture closely follows our foundational LLM: a decoder‑only Transformer trained with next‑token‑prediction style objectives, augmented with a catalog‑aware ranking head that scores only Netflix in-catalog items. The scoring pipeline works as follows:

1. Verbalization: A verbalizer V serializes user history H, context 𝜏 , and relevant item metadata into a single text sequence x.
2. Pooled representation: The LLM processes x, and we extract a pooled hidden state h that summarizes the user’s current preferences and context.
3. Catalog‑aware scoring: Each catalog item i has a learned embedding eᵢ. A scoring head ϕ combines h and eᵢ (e.g., via dot product or small MLP) to produce a score sᵢ. Applying a softmax over scores yields a probability distribution which we convert into a ranking π.

All parameters — the backbone, scoring head, and item embeddings — are trained jointly. For very large catalogs, we can use sampled softmax or candidate sets for efficient training and inference. This architecture constrains recommendations to the Netflix catalog while supporting efficient scoring over large candidate sets.

Serving and Cost Optimization

GenRec is served on Netflix’s internal LLM stack using vLLM. At Netflix scale, serving cost is driven primarily by 1) Model size; 2) Context length; 3) Inference mode (prefill vs. autoregressive decoding). We control cost through three strategies:

• Smaller / distilled models: We train GenRec on smaller or distilled foundation models, often with larger or more targeted datasets, to capture most of the quality of larger models at lower serving cost.
• Aggressive context compaction: Using the context engineering described earlier, we minimize tokens while preserving ranking quality.
• Prefill‑only inference: Autoregressive decoding over large candidate sets would be prohibitively expensive. Instead, we run in prefill‑only mode: the model consumes the prompt once and scores the entire candidate set in a single forward pass, with no token‑by‑token decoding.

Together, these choices make it feasible to serve GenRec on high‑volume workloads within compute budgets.

Offline and Online Experiments

We evaluated GenRec against a mature production ranker that has been tuned over many years. The baseline model relies on thousands of engineered dense and embedding features, as well as custom architectures for modeling feature interactions and sequences. We assessed performance using both offline evaluation metrics and a large‑scale online A/B test.

GenRec vs Production Baseline

Offline, GenRec outperformed the production ranker on ranking metrics despite using far fewer input signals and labeled examples. With roughly 40× fewer Phase‑2 labeled training examples, GenRec achieved about +1.6% improvement in Mean Reciprocal Rank (MRR). As we increased Phase‑2 training data and enriched the input signals, GenRec’s offline metrics continued to improve.

Online, we ran a large A/B test on batch‑compute recommendation surfaces, covering ~10% of Netflix traffic over ~4 weeks. In this low‑data, low‑signal configuration, GenRec delivered statistically significant gains over the production baseline on both short‑term and long‑term online metrics (Figure 3).

These results indicate that a properly post‑trained and aligned LLM‑backed ranker can be a strong alternative to traditional recommendation models, with substantial headroom as we further scale data and input signals.

Data, Model, and Phase Contributions

We ran ablations to understand where GenRec’s gains come from.

Data and Model Scaling

• Data scaling: For both ~1B and ~10B parameter backbones, offline MRR improves as we increase Phase‑2 post‑training data. Larger models reach higher absolute MRR but follow a similar scaling curve (see Figure 4).
• Model scaling: Under a fixed training budget, we post‑trained GenRec variants from ~1B to ~10B parameters. Within this budget, larger backbones consistently achieved higher offline MRR than smaller ones.

Phase-1 vs. OSS, Phase-2 vs. Phase-1

• Phase-1 vs. OSS: Using the Phase‑1 Netflix‑adapted foundation LLM as the base model improves offline ranking metrics by roughly 10–20% compared to starting directly from an off‑the‑shelf LLM.
• Phase-2 vs. Phase-1: Phase‑2 post‑training adds another 35–50% gain in offline ranking metrics when evaluated near the Phase‑1 training cutoff (i.e. when Phase‑1 model is the freshest). As time passes and Phase‑1 becomes stale with new content and shifting tastes, the relative benefit of Phase 2 grows to about 80% after 2 weeks.

Data efficiency vs. production ranker

• Starting from a strong Phase‑1 model, GenRec matches or exceeds the production ranker using 10–40× fewer Phase‑2 labeled examples, depending on configuration. This marginal data efficiency is especially valuable because Phase 2 is refreshed far more frequently than Phase 1.

Context Length Optimization

Context length drives both quality and cost: longer verbalizations expose more behavior and context but increase training and serving cost. To study this trade‑off, we varied context length and verbosity and optimized them in three steps:

1. Clean and compress events: drop low‑signal engagements and compress repetitive behavior to form a cleaned sequence of events.
2. Find the “elbow point”: vary how many historical events we include and plot MRR vs. number of events to identify an elbow beyond which additional context yields diminishing returns (see Figure 5).
3. Optimize verbosity: for the retained events, test different levels of details and simplified wordings, measuring MRR each time.

In our experiments, we can reduce the context tokens to roughly one-third of the original budget with negligible degradation in offline ranking metrics. Since serving cost is approximately proportional to context length, we observed a similar reduction in serving cost.

Towards LLM‑Native Recommendation

GenRec is more than “swapping in a Transformer” for an existing ranker. It hints at a broader shift toward LLM‑native recommendation at Netflix. A few notable changes:

From Feature Engineering to Context Engineering

Traditional RecSys stacks revolve around large feature sets and heavy feature infrastructure. LLM‑centric systems instead revolve around constructing rich textual contexts from raw logs, metadata, and tools. The “prompt” becomes the new feature vector.

Modeling effort shifts from designing features to deciding which signals to include, how far back in time to go, how to compress or summarize history within a token budget. Our experiments on verbalization compaction illustrate this shift: careful context design can preserve quality while dramatically reducing serving cost.

From Customized Architectures to Foundation Backbones

Historically, each recommendation task often had its own custom architecture (two‑tower models, DLRM‑style networks, bespoke attention blocks). In an LLM‑centric world, many tasks share a common foundation backbone, with differentiation coming from data and verbalization strategies, post‑training objectives and rewards, and inference optimization.

GenRec leverages the same backbone as our foundation LLM rather than introducing a new architecture built from scratch. This makes it easier to share learnings across applications, and opens the door to natural‑language steering for future experiences.

Scaling Laws as Design Guides

Traditional RecSys can hit diminishing returns due to sparse IDs, heavy engineering objectives, and task‑specific architectures. With an LLM‑backed backbone, recommendation inherits clearer data and model scaling behavior: within cost limits, more data and larger models consistently improve quality. This brings RecSys design closer to the broader LLM paradigm, where scaling laws help guide model and data investment.

From RecSys Infra to LLM Infra

LLM‑backed recommenders push us toward LLM‑style infrastructure: GPU‑accelerated, vLLM/Triton‑based, with careful batching and caching. Over time, recommendation serving infra starts to look more like general LLM infra than classic RecSys stacks built around MLPs or factorization models.

Conclusions

We have presented GenRec, an LLM‑backed recommendation ranker at Netflix that adapts an internal foundation LLM for large‑scale personalization. By verbalizing user histories, context, and item metadata, adding a catalog‑aware ranking head, using reward‑weighted objectives aligned to long‑term satisfaction and business goals, and serving efficiently on our LLM infrastructure, we obtain a model that improves on a strong production ranker while using far fewer Phase‑2 labels and input signals.

GenRec is an early but promising step toward a more LLM‑centric recommendation stack at Netflix. Our results suggest that, with careful attention to cost, infrastructure, and alignment, LLM‑backed recommenders can play a central role in large‑scale personalization.

Acknowledgments

GenRec is the result of close collaboration among multiple teams and organizations across Netflix. The contributors to this work (in alphabetical order):

AI for members: Arjun Rao, Ashish Rastogi, Baolin Li, Fernando Amat Gil, Grace Huang, Justin Basilico, Kamelia Aryafar, Linas Baltrunas, Moumita Bhattacharya, Ogheneovo Dibie, Rein Houthooft, Shradha Sehgal, Sejoon Oh, Sergi Perez, Sourabh Medapati, Thea Wang, Yaochen Zhu, Yesu Feng, Ying Li, Yun Li, Yucheng Shi, Yunan Hu

AI platform and serving: Abhishek Agrawal, Adam Singer, Binh Tang, Daneo Zhang, Derek Olejnik, Ed Maddox, Erik Osheim, Lingyi Liu, Liping Peng, Meghana Chilukuri, Nicolas Hortiguera, Shaojing Li, ZQ Zhang

Product: Ilke Kaya, Michelle Kislak, Scarlet Chen, Si Cheng

GenRec: Towards LLM-Native Recommendation at Netflix was originally published in Netflix TechBlog on Medium, where people are continuing the conversation by highlighting and responding to this story.

This stylish biography of the author traces his journey from left to right, and from ‘posh Lima’ to Cricklewood

“At what precise moment had Peru fucked itself up?” Those words on the opening page of Conversation in the Cathedral announced the arrival of a writer who understood that national history could become the organising principle of fiction. At precisely the moment when Latin America was discovering the intoxicating possibilities of magical realism – a form whose exuberance had an elective affinity with revolutionary politics, collapsing the distinction between the marvellous and the possible – Mario Vargas Llosa plumped for old-fashioned realism instead.

It was a question of sensibility as much as politics, Gerald Martin suggests in this chatty yet coolly analytical biography. He traces the path of a novelist who began by dissecting oligarchy with surgical precision and ended happily dining among its ranks, completing one of literary history’s most spectacular political migrations – from the hard left to the hard right, over the course of half a century.

Composers from Purcell to Birtwistle have created passionate dramas inspired by ancient Greek stories – and none of them require you to sit through three hours of solemn spectacle without an interval

For me, the most astonishing thing about Christopher Nolan’s The Odyssey is its barnstorming popularity. I found this cinematic epic for the ages visually bloodless, expressively monotonous and very, very long. Mind you, it is good to know that Nolan thinks that war is, on balance, you know, once you really think about it, probably a bad idea.

The music, too, was disappointing. If I were Nolan, I’d have asked Mica Levi to score The Odyssey. Ludwig Göransson could have surely done more to make a soundworld of genuine strangeness, but all he could manage is a falling tritone and then a descending fifth on the two reeds and pipes of a recreated aulos, which he uses as a melancholy leitmotif in all-too-conventional Hollywood fashion. (The aulos is beautifully played by Callum Armstrong on the soundtrack, at least.) The best sounds in the film come in the scene with Samantha Morton’s Circe, in which the crew’s gluttony is horrorshow ASMR: surround-sound mastication that threatens to devour the audience.

Buratto/De Tommaso/Orchestra I Pomeriggi Musicali/Milletarì
(Decca)
The long-lost score, finally premiered 138 years after it was written, brims with melodic promise despite its dramatic shortcomings

It was a bold move by the 21-year-old Umberto Giordano to enter the 1888 one-act opera competition won by Mascagni’s Cavalleria Rusticana. Marina, set in Montenegro during an unspecified conflict with Serbia, came sixth, the score languishing until 2026, when it received its belated premiere at Milan’s Teatro dal Verme. This recording reveals a composer with a keen ear for musical colour even if the opera is no masterpiece.

When Marina, a Montenegrin, takes in Giorgio, a wounded Serb, the inevitable conflict with her brother and her bullish suitor precipitates a typical verismo catastrophe. After a militaristic opening, Marina takes centre stage with an aching ballad of lost opportunity. Her scene with the injured Giorgio follows abruptly, redeemed by Giordano’s knack for trotting out appealing melodies. It’s not the most sophisticated writing but it comes straight from the heart.

The multi-instrumentalist faced resistance when he premiered his sound in the 60s but audiences have fallen in love with it since. Now aged 82, he’s still finding new directions – and wants recognition for Africa’s indigenous artists

“Do you know the Metro club?” I feign recognition to the 82-year-old father of Ethio-jazz, Mulatu Astatke, who knows the UK capital’s vanished haunts better than I do (the Soho venue closed down decades ago). By then, he had immersed himself in the city’s 60s jazz scene, befriending Ronnie Scott and the saxophonist Joe Harriott, with whom he occasionally jammed across Soho’s clubs, as well as musicians from Africa and the Caribbean.

But when he returned to “Swinging Addis” years later with the new genre he’d innovated in England and the United States, the audience at one of his first performances weren’t so taken by his fusion of jazz riffs with the harp-like begena. “Stop,” Astatke was told. “No more!” He didn’t say a word. Instead, he bowed, said his thanks and walked off. “I said: ‘I’ll show you,’” recalls Astatke in his smoky voice. “‘One day, I’ll come back and show you.’”

Louise Brealey teases out the playful and poetic moments of these shapeshifting stories, told from the perspective of the northeasterly wind and those it has wreaked chaos on through the ages

The only named wind in the British Isles, Helm is a strong northeasterly gale that blows down the slopes of Cross Fell in the Pennines and is known for its unusual cloud formations. Helm has been known to dislodge roofs, haystacks, trees and riders from their horses.

Twenty years in the writing, Sarah Hall’s Helm is an ambitious novel that tells stories from the perspective of the wind, and from the people on whom, through the ages, Helm has wreaked chaos. Those figures include a Neolithic tribeswoman and seer called NaNay who is building a stone circle; an astrologer and exorcist known as a “wizard priest”; a Victorian meteorologist who travels across the Pennines by train and later by carriage into Helm’s “onslaught”; and a climatologist measuring microplastics in the atmosphere from her observation hut in the fells. Separated by centuries, their stories are nonetheless woven together, each of them bonded by weather.

Late-night hosts speak about Trump’s trip to Michigan, where a local again shouted the same name at him

Late-night hosts spoke about Donald Trump’s recent Michigan visit and his attendance at Lindsey Graham’s funeral.

Dominic Sessa stars as a conceited yet vulnerable young chef alongside Antonio Banderas – but the feelgood factor is compromised by what we know of the author

For all the drugs and punch-ups, this is an unexpectedly sentimental coming-of-age story – or maybe origin myth – about the formative youthful kitchen experiences of legendary badass chef Anthony Bourdain. It is taken from the opening chapters in his bestselling memoir Kitchen Confidential; Canadian film-maker Matt Johnson directs and Dominic Sessa (the callow kid from Alexander Payne’s The Holdovers) plays the conceited, smug yet lonely and vulnerable young college student Tony. “Anthony” is in the future.

It is the mid-1970s and Tony effectively runs away from home after failing to get a residential summer writing scholarship that he had assured everyone was in the bag. He shows up nearly penniless in the seaside vacation resort of Provincetown, Massachusetts – a place that is pretty similar to Amity Island in Jaws, which is showing at every drive-in. The reason is that he has a desperate crush on Nancy (Emilia Jones) who is there for the summer too and whom he desperately wants to impress with his smouldering intellectualism.

Four women speak out about their underage experiences with the movie star and musician. Are the floodgates finally about to open?

Jared Leto has long been a person with something of a reputational issue. The public-facing side of this was bad enough – while playing The Joker in 2016’s Suicide Squad, he reportedly sent rats, dead pigs and used condoms to his cast mates – but for years there have been upsetting rumours about what he does in private. After a lengthy investigation, the BBC documentary Jared Leto: Hollywood’s Dark Secret has finally brought at least some of them to light.

This is, as you might imagine, not a fun thing to watch. The film, directed by Alice McShane, documents allegations of what are said to be decades of systematic criminal misconduct against teenage girls. The women interviewed here allege sexual assault, threatened sexual assault, sexualised grooming and statutory rape. They are backed up by anonymous male former co-workers who claim to have witnessed the machinery that Leto set up to procure underage girls from concerts.

The prolific crime author’s latest Netflix thriller I Will Find You has become one of the streamer’s most-watched shows of all time – but what’s his secret?

Netflix’s all-time top 10 English language series chart is starting to look like a stale old thing. It is topped by Wednesday, which is four years old, and includes dusty old hits including Stranger Things season four (four years old), Monster: The Jeffrey Dahmer Story (four years old), The Queen’s Gambit (six years old), and Bridgerton season one (six years old).

But wait. Just before you write off Netflix as a platform half a decade past its peak, a plucky new entry has entered the all-time top 10 to liven things up. That show is I Will Find You, the latest belched out from the glistening ovipositor of Harlan Coben. If you’re a fan of Coben’s work – and we’ll get into that in due course – then I Will Find You will be instantly recognisable, since it bears many of the author’s hallmarks.

Deep underground the player joins a colony of ruthlessly exploited pit workers who could be taking on the bosses. How things play out will depend on you

It is rare for video games to feature industrial action as a key theme, but it does happen. There’s the shady Dockworker’s Union in Disco Elysium and the Orbital Workers Union in space adventure Tacoma; then there is Lynx, the evil union-bashing corp in Hardspace: Shipbreaker. Soon we can add A Line Held Tight, a visual novel that puts workers’ rights – and workers’ lives – at the very centre of the narrative.

In the distant future, a colony of miners digs far beneath the surface of an isolated planet, searching for veins of some exotic ore, protected from the poisonous atmosphere within giant mech suits. But the working conditions are woeful and as a strike looms, an unpopular employee named the Canary must decide whether or not to stand with colleagues and fight against suppression. The player, of course, takes on the role of the Canary, navigating personal as well as political struggles in a world rendered in stark monochrome.

(American Dogwood/Empire)
Beneath the Hollywood narrators, Pink Floyd-y interludes and accompanying comic books lies songwriting that proves the Tipsy singer’s breakout success was no fluke

It’s fair to say that if Shaboozey never released another note of music, his place in history would still be assured.

He wasn’t the only Black country-leaning artist brought to a wider audience by a guest appearance on Beyoncé’s 2024 album Cowboy Carter, but no one else capitalised on the exposure quite so remarkably. Released two weeks after his guest appearance on that album’s Spaghettii, his single A Bar Song (Tipsy) unexpectedly eclipsed Beyoncé’s success, becoming the biggest-selling country single of the year – more than 1.75bn streams and counting on Spotify – and for a time, the joint longest-running US No 1 of all time. He’s received three nominations at the Country Music awards, an event traditionally resistant to artists of colour. The accompanying album, Where I’ve Been, Isn’t Where I’m Going, was by far the most successful sustained attempt to fuse hip-hop and country, two genres that have been warily circling each other for some time, spawning the occasional novelty hit such as Nelly and Tim McGraw’s Over and Over Again, the Ludacris-featuring 2010 remix of Jason Aldean’s Dirt Road, and Lil Nas X’s Old Town Road; the baleful, rap-influenced spectre of bro country; and the deep underground sub-genre of hick-hop, a hidden world of Maga politics, confederate flags and festivals held at Georgia mud holes. The album is still on the US chart, 112 weeks after its release.

Environmentalists renew calls for windfall tax to support households after Q2 earnings rise to almost $10bn

Shell has posted its second highest quarterly earnings on record as the energy market surge caused by the Middle East crisis helped to double its net profit to almost $10bn in the three months to June.

Europe’s biggest oil and gas company reported a quarterly profit of $9.8bn (£7.4bn), more than double the figure for the second quarter last year and the highest since its record-high earnings in the months after Russia’s invasion of Ukraine.

With his popularity at a record low, Friedrich Merz packs the papal guide Magifica humanitas

• Europe live – latest updates

Summer is normally thought of as the time for pacy potboilers and brain-soothing page-turners – airport reads that wind up sandy and sunscreen-stained. Not so for Friedrich Merz’s holiday reading list.

Germany’s embattled chancellor, who has faced one rough patch after another since taking office 14 months ago, seems to be seeking spiritual guidance this year while putting his feet up.