Newsletter platform Beehiiv is expanding into new avenues of engagement by launching a feature called Community, which lets subscribers of a creator chat with each other. The company also launched a new AI Copilot that helps creators manage and grow their audience.

The updates come as Beehiiv positions itself as a creator platform beyond newsletters. In the last few months, the company has launched podcasts, webinars, and customizable paywalls. Some of these moves are already showing positive results. The company said that 50% of podcast users migrated their shows from elsewhere, for instance.

Beehiiv’s new Community tool will allow users to spin up a discussion forum within the platform. Today, creators often have a chat for members on a separate Discord or Slack server or in Facebook groups, but Beehiiv wants to bring those chats back to its own platform. Here, creators can also create paid membership tiers for exclusive access to certain chatrooms and moderate conversations.

“People following your content have a shared interest in what you’re creating, but they can’t communicate with each other. Whether that interest is in sports, the World Cup, or politics, being able to have a community where your audience can actually engage with one another is super valuable,” Beehiiv CEO Tyler Denk told TechCrunch.

The platform is also introducing an additional revenue-generation opportunity with programmatic ads, which allow users to sell ad slots in their newsletters. They can earn money by choosing the ads that potentially offer the highest returns based on their audience, content, and performance.

The company already has tools like metered paywalls, paid trials, and a sponsorship storefront to sell their own slots in packages. Plus, Beehiiv said the publishers on the platform earn more than $1 million per month through their ad network.

Beehiiv is launching a new AI assistant called Copilot, as well, which can understand context like content, audience, subscribers, and performance to give users advice on how to manage their newsletter and grow their audience. The assistant can analyze the performance of various newsletters and podcasts, draft campaigns for outreach, and look for new money-making opportunities.

The assistant is one of several AI efforts underway. Earlier this year, the company launched a model context protocol (MCP) server, allowing users to connect their Beehiiv to other assistants like ChatGPT and Claude to ask questions and get insights. It’s also working on better AEO (Answer Engine Optimization), which helps a newsletter be cited in AI assistant answers more frequently.

Along with these updates, the company is shipping a redesigned editor that allows users to see editing and preview modes side by side, helping them to understand how the content they are writing would appear to readers.

Denk noted that in the coming quarter, Beehiiv wants to spend time educating users about these tools and teaching them about how top newsletters are using them to grow their publications.

The platform’s rivals are also evolving by launching new offerings. For instance, Riverside launched a newsletter publishing feature last month, and Substack launched a built-in recording studio product in March.

Google is expected to launch the Pixel 11 series at its August 12 event. Ahead of the official unveiling, a leak has apparently exposed the prices of the new smartphones in the US, along with their storage capacity, colors, and specifications. These alleged leaks were discovered from Amazon listings that were taken down. As these details haven’t been officially verified yet, they remain just rumors.

Pixel 11 Series Expected Specifications

The leaked specifications from Android Authority indicate Google may keep several familiar hardware features in the Pixel 11 lineup. The standard Pixel 11 could offer a 6.3-inch OLED display with a 120Hz refresh rate. It could also come with a 4,985mAh battery, a 13MP front-facing camera, Bluetooth 6, and Wi-Fi 6E capability.

Pro models will also get some hardware modifications. For instance, the Pixel 11 Pro is likely to have a 4,850 mAh battery, whereas the Pro XL model can have a 5,115 mAh battery. Both devices may feature 13MP front cameras. According to rumors, the 256 GB variant will have 12 GB of RAM, while other storage options will have 16 GB of RAM. The Pixel 11 Pro Fold could feature a 6.5-inch OLED cover display, a 4,750mAh battery, and a 13MP front camera.

The leaked information further suggests Google may revise both pricing and storage with the Pixel 11 series. Every model could see a $100 price increase. However, the base Pixel 11 may compensate with 256GB of storage as standard. The Pro models could also receive a different RAM setup. The base 256GB variants may feature 12GB RAM. Higher storage options are expected to come with 16GB RAM.

Leaked Pricing, Storage Options, and Colors

The lineup is expected to include four models: Pixel 11, Pixel 11 Pro, Pixel 11 Pro XL, and Pixel 11 Pro Fold. It also hints that the base Pixel 11 may no longer offer a 128GB storage option.

The UK looked set to crack down on VPNs as it tightens the rules for children online. Instead it has backed off, and its own research is the reason why.

Online Safety Minister Kanishka Narayan put it plainly on the BBC. “We decided not to limit VPNs,” he said. A VPN hides a user’s real location, which is one way to slip past an age check.

The decision landed alongside the UK’s new midnight social-media curfew for 16 and 17-year-olds. Technology Secretary Liz Kendall confirmed it in a written statement, saying VPNs have “legitimate privacy and security uses.”

What the research found

The government had commissioned a study of more than 2,000 children, and the numbers undercut the case for a ban. About a quarter of 11 to 17-year-olds have used a VPN. Most do so for privacy, not to break the rules, the report found.

Only around 7% of children use a VPN to reach age-restricted content. Far more simply lie. Nearly half who dodge an age check just enter a false date of birth. The VPN, in other words, is not the main loophole.

The burden shifts to platforms

Rather than police the tools, the government is pushing the job onto the platforms. They must now take “robust steps” to spot and stop under-age users getting around age checks.

Ofcom must report by October on what a robust over-16 age check looks like. The government has separately asked it, with the data regulator, to study how platforms can better detect VPN use. Ministers will also talk to VPN providers about voluntary action, and say they will “keep this area under close review.”

A win for privacy campaigners

The retreat is a clear win for digital-rights groups. A coalition of more than 20 tech firms and campaigners, including Proton and Mozilla, had urged ministers to leave VPNs alone. Mozilla warned that age-gating them would create a cybersecurity mess while failing to protect children.

Not everyone thinks the wider plan works. The curfew and the feature limits can be switched off, and critics say that leaves an obvious gap. The government is “leaving the side door open,” as one analyst put it.

The move stands out against the global mood. It sits alongside the UK’s coming under-16 social-media ban, while Australia’s teen ban has been dogged by VPN workarounds and New Zealand recently ruled out its own limits. Even the US courts are wrestling with who runs the internet’s age gate. For now, Britain has chosen evidence over a blanket ban.

GIGABYTE has officially entered India’s local manufacturing ecosystem with the launch of the GAMING A16, the company’s first Made-in-India gaming laptop. The new laptop is powered by AMD Ryzen processors and is designed for gamers, creators, students, and professionals seeking AI-ready performance. The launch also marks GIGABYTE’s first gaming laptop to be manufactured in India, with production handled in partnership with Dixon Technologies, one of the country’s largest electronics manufacturers.

According to the company, the move is part of its long-term strategy to strengthen its presence in India while supporting the country’s push toward becoming a global electronics manufacturing hub. GIGABYTE says local production will also help improve supply chain efficiency and allow it to respond more quickly to market demand.

To oversee the project, senior executives from GIGABYTE’s global headquarters—including teams from product management, procurement, materials planning, and quality assurance—visited India to work alongside Dixon during production. The company says every unit has been built to meet the same quality standards as its globally manufactured products.

Designed for Gaming and AI Workloads

While GIGABYTE hasn’t shared the complete hardware specifications in today’s announcement, it confirmed that the GAMING A16 is powered by AMD Ryzen processors and is designed to handle both modern gaming and AI-assisted workloads.

The company says the laptop targets a wide audience, including gamers, content creators, college students, and professionals who need a balance of performance and portability. The launch also reflects the growing importance of AI-ready PCs, with GIGABYTE positioning the A16 as a machine capable of handling both next-generation AI experiences and traditional gaming workloads.

Speaking on the matter, Sinclair Hsiao, Vice President of Global Sales at GIGABYTE, said,

India has become one of the most exciting technology and gaming markets in the world, and we believe Indian consumers deserve products built to the same uncompromising global standards that gamers everywhere expect from GIGABYTE

The GIGABYTE GAMING A16 is now available in India. While the company has confirmed local availability, it has yet to announce pricing or detailed hardware configurations

Google might launch an AI product with one name during its experimental phase, but it will eventually tie it all to Gemini. In the latest example of this trend, the company is renaming its AI-powered research product NotebookLM to Gemini Notebook. The company is also adding features to make the tool more interactive by infusing coding execution for data analysis.

The company first showed off NotebookLM during Google IO in 2023 as Project Tailwind, and since then, it has made it into a product used by 30 million people and over 600,000 organizations. In the last three years, the company has added capabilities, like interactive podcast generation, curated notebooks, video overviews, support for more file types, and an enterprise plan.

Because of NotebookLM, other companies and startups have added capabilities for podcast generation from source material and research tools.

Along with renaming, Google is rolling out a new update that makes each notebook its own secure container, in which users can generate code to make outputs interactive. It noted that with code execution ability, users can tap into multiple sources and create complex data analysis directly within the tool.

The company said the update is available to Google AI Ultra paid plan users, along with Workspace business customers with AI Ultra Access and AI Expanded Access. Pro users will get access to this feature in the coming weeks.

Google said that users can already look at their notebooks within the Gemini app, and soon, they will be able to access them through AI Mode in search.

Enterprises looking to move more of their agentic AI workloads to open weights models they can customize, control and run on-premises or in virtual private clouds have a strong new contender to consider.

Today, Thinking Machines—the highly capitalized American AI startup founded by former OpenAI CTO Mira Murati—released Inkling, its first major language model under an enterprise-friendly Apache 2.0 open source license, and it boasts high, if sub state-of-the-art, performance for open weights models on third-party benchmarks, specifically software engineering (77.6% on SWE-bench Verified, where it beats fellow U.S. open rival Nvidia Nemotron 3’s 71.9%) and voice understanding (91.4% on VoiceBench compared to 94.4% for Gemini 3.1 Pro on high reasoning effort).

Another differentiator: Thinking Machines notes that Inkling was designed “to answer directly on topics that may be subject to censorship,” offering enterprises concerned about factual outputs, irrespective of controversy or sensitivity, a more trustworthy option.

Coming in at 975 billion total parameters, Inkling is a natively multimodal, open-weights Mixture-of-Experts (MoE) system capable of reasoning across text, images, and audio. The weights are already available on Hugging Face and the company’s own model training application programming interface (API), Tinker.

Designed to balance cost against performance through a novel “controllable thinking effort” mechanism, the model represents a significant departure from the black-box scaling strategies of frontier competitors.

Alongside the flagship model, Thinking Machines also announced a preview of Inkling-Small, a lighter 276-billion-parameter alternative optimized for workloads where low latency and cost are paramount.

Benchmarks Show a Powerful, High-End, Sub State-of-the-Art Model

While Inkling is a formidable multimodal engine, it lands in a fiercely competitive 2026 open-weight landscape characterized by highly specialized MoE architectures. Rather than attempting to dominate every leaderboard, Thinking Machines explicitly designed Inkling—with 975 billion total and 41 billion active parameters—as a broad, balanced generalist.

For example, it comes in near the middle high-end of benchmark performance 1257 on Design Arena’s Agentic Web Dev leaderboard measuring human scores of frontend web design.

But China’s leading AI labs have produced models with elite reasoning and coding capabilities, posing a stiff challenge to Inkling’s generalist approach and ultimately outperforming it on general and coding benchmarks.

• GLM 5.2: Widely considered the top open-weight reasoning model available in the benchmark set, GLM 5.2 outperforms Inkling on pure coding, agentic, and complex reasoning tasks. It scores 62.1% on SWEBench Pro (Public) compared to Inkling’s 54.3%, and a massive 82.7 on Terminal Bench 2.1 against Inkling’s 63.8. GLM 5.2 also holds the edge in text-only reasoning, scoring 40.1% on HLE (text only) versus Inkling’s 30.0%.

• DeepSeek V4 Pro: DeepSeek maintains an edge in several strict coding and factuality domains, beating Inkling on SWEBench Verified (80.6% vs. 77.6%) and SimpleQA Verified (57.0% vs. 43.9%). However, Inkling successfully overtakes DeepSeek V4 Pro in mathematical problem-solving, achieving 97.1% on AIME 2026 compared to DeepSeek’s 96.7%.

• Kimi K2.6: This model outpaces Inkling across multiple technical benchmarks, delivering higher scores on GPQA Diamond (91.1% vs. 87.9%), BrowseComp (83.2% vs. 77.1%), and HLE with tools (54.0% vs. 46.0%). Yet Inkling proves more resilient on general chat instruction following, scoring 79.8% on IFBench compared to Kimi K2.6’s 76.0%.

Against its primary U.S.-based open-weight competition, Inkling demonstrates strong parity and frequent superiority.

• Nemotron 3 Ultra: Inkling consistently outperforms this U.S. rival across reasoning and coding. Inkling posts 97.1% on AIME 2026 and 77.6% on SWEBench Verified, beating Nemotron’s 94.2% and 70.7%, respectively. Furthermore, Inkling significantly leads in agentic workflows, scoring 74.1% on MCP Atlas against Nemotron’s 44.7%.

When compared to closed-source juggernauts like Claude Fable 5, GPT 5.6 Sol, and Gemini 3.1 Pro, Inkling trails in peak reasoning and software engineering autonomy, but remains highly competitive in multimodality.

• Coding and Reasoning: Closed models maintain a commanding lead. Claude Fable 5 (max) hits 95.0% on SWEBench Verified and 53.3% on HLE (text only), far outpacing Inkling’s 77.6% and 30.0%. GPT 5.6 Sol dominates Terminal Bench 2.1 with an 89.5, easily clearing Inkling’s 63.8.

• Native Multimodality: Inkling’s native visual and audio capabilities hold their own. On the MMMU Pro (Standard 10) vision benchmark, Inkling’s 73.3% is competitive, though trailing Claude Fable 5’s 84.2% and GPT 5.6 Sol’s 83.0%. In audio processing, Inkling scores a highly respectable 77.2% on MMAU, keeping it within striking distance of Gemini 3.1 Pro’s 82.5%.

If an enterprise workflow demands elite software engineering autonomy or the highest bounds of text-only reasoning, models like GLM 5.2 or proprietary systems like Claude Fable 5 maintain the edge.

However, Inkling carves out a unique and highly defensible position: it is the most capable open-weight foundation model that natively fuses text, vision, and audio, while simultaneously offering developers direct programmatic control over the cost-to-performance ratio.

The Shift from Static Reasoning to Controllable Thinking

Rather than attempting to build a singular “god model” optimized strictly for state-of-the-art benchmark domination, Thinking Machines engineered Inkling for adaptability and efficiency in real-world workflows.

The standout feature of this release is Inkling’s “controllable thinking effort.” Developers can programmatically adjust the model’s reasoning budget—scaling from 0.2 to 0.99—to dictate how hard the AI should “think” before generating an output.

As the company noted, “Inkling’s continuous thinking effort lets you pick your point on the cost/performance curve—reaching the same score with a fraction of the tokens”.

In practical terms, this allows enterprises to deploy Inkling with lower token expenditure for simpler tasks, while cranking up the compute overhead for complex, multi-step reasoning challenges. However, by keeping the thinking effort lower and generating fewer tokens, the cost-conscious enterprise can achieve high quality results and performance on simple tasks while spending less money, or, in the case of those running models locally, less costs on energy and compute resources.

During the model’s large-scale reinforcement learning (RL) training over 30 million rollouts, researchers observed an emergent phenomenon they called “chain of thought condensation”. Over time, Inkling naturally learned to compress its internal reasoning steps—dropping grammatical overhead and connectives—while reaching the same accurate conclusions, resulting in drastically reduced latency.

Epistemics and Censorship Resistance

A notable element of Thinking Machines’ release is its explicit focus on the model’s epistemics—specifically its calibration, instruction following, and resistance to censorship.

In an ecosystem where open-weight models adopt either overly restrictive safety guardrails or echo state-aligned ideological talking points, Inkling was intentionally trained to answer directly on politically sensitive or heavily censored topics.

To validate this approach, Thinking Machines submitted Inkling to the Propaganda and Censorship Eval developed by AI startup Cognition. According to the published findings, Inkling demonstrated “strong patterns of censorship non-compliance,” effectively resisting ideological capture or boilerplate refusals when presented with sensitive subjects.

Despite its resistance to censorship, the model maintains a robust defense against genuinely malicious, dangerous, or illegal queries. On the StrongREJECT benchmark—which tests responses to unambiguous harmful requests—Inkling scored 98.6%, placing it in line with strict frontier safety standards. Furthermore, on the FORTRESS benchmark, Inkling successfully navigated the line between safety and over-refusal: it achieved a 78.0% refusal rate on adversarial queries (such as those involving weapons, cyberattacks, or violence) while maintaining a 95.9% compliance rate on benign, look-alike queries.

Thinking Machines noted that typical open-weight vulnerabilities remain within the architecture. Internal safety evaluations revealed an “occasional tendency to comply with role-play and indirectly framed prompts concerning harmful topics”. The company advised enterprise developers to treat the model’s built-in refusals as just one layer of security, recommending the downstream deployment of external moderation tools—such as Llama Guard—to filter adversarial jailbreaks and enforce use-case-specific safety policies at the application level.

Under the Hood: Architecture and Multimodality

Inkling’s scale is staggering, yet sparse. The MoE architecture features 975 billion total parameters, but only 41 billion parameters are active during any given token generation. It supports a massive context window of 1 million tokens and diverges from typical transformer models by using relative positional embeddings instead of the industry-standard Rotary Positional Embedding (RoPE).

True to the company’s foundational vision, Inkling was trained from scratch to be natively multimodal. Unlike models that rely on bolted-on external encoders, Inkling uses an encoder-free early fusion approach. It directly ingests audio as discrete dMel spectrograms and visual data as 40×40 pixel patches via a hierarchical multi-layer perceptron (hMLP), projecting all modalities into a shared hidden space.

Licensing: True Open-Source for the Enterprise

For enterprise IT teams and developers, the most disruptive aspect of Inkling may be its licensing. Inkling is released under the permissive Apache 2.0 license.

In an ecosystem where many so-called “open” models from Western labs are tethered to dual-use commercial licenses, acceptable use restrictions, or revenue caps, an Apache 2.0 designation makes Inkling a true open-source foundation. This gives developers the legal freedom to download, modify, integrate, and commercialize the model weights entirely royalty-free.

The model is readily deployable across major open-source inference libraries—including SGLang, vLLM, TokenSpeed, and llama.cpp—and comes with a native NVFP4 quantized checkpoint optimized for NVIDIA Blackwell systems.

Community Reactions: The Engineering Feat

The AI community’s response has been swift, praising both the model’s openness and the underlying engineering execution.

In a post on X, Thinking Machines co-founder John Schulman reflected on the rapid development cycle: “Inkling is out today, with open weights and in Tinker. It’s been fun to watch this one come together: pretraining began last winter, and starting in mid-January a small team built up the coding, reasoning, and agentic training from there. We learned a lot building it, and I hope people find good uses for it.”

Horace He, a researcher at Thinking Machines (previously from PyTorch), underscored the difficulty of the task in another post on X: “It truly takes a village to release a model, perhaps especially an open weights model. Actually doing the entire process from scratch, from data to pretraining to posttraining to actual release, gives a lot of appreciation for anyone who does it!”

The broader open-source ecosystem has also embraced the technical integrations. Lysandre Debut, the Chief Open-Source Officer at Hugging Face, shared his enthusiasm regarding the model’s optimization in his own X post: “One thing I find quite striking is how much easier accelerating models has become… We replaced the model’s causal Conv1D with the `causal-conv1d` kernel. One line changed, +4% tokens per second. We then replaced its attention implementation with FlashAttention-4. Another single change, another +11%. That’s a total throughput improvement of about 15%, without changing the model architecture or retraining anything.”

Tiezhen Wang, an ecosystem growth expert and ex-Googler, celebrated the release as a massive win for the open-source community, listing the model’s impressive specifications on X, highlighting its “975B total, 41B active” size, “Native MTP support,” and the highly coveted “Apache 2.0 license.”

Background: The Road to Inkling

To understand the significance of Inkling, one has to look back at the rapid trajectory of Thinking Machines over the past 18 months.

When Mira Murati departed OpenAI in late 2024 to found Thinking Machines alongside industry veterans like John Schulman and Barret Zoph, the stated goal was to pivot away from building isolated autonomous agents. Instead, the company aimed to build flexible, multimodal systems designed for genuine human-AI collaboration and open science.

By July 2025, the startup had secured a historic $2 billion seed round led by Andreessen Horowitz at a $12 billion valuation. At the time, Murati promised the impending release of a product with a “significant open source component” to empower researchers and startups.

The company’s philosophy began coming into sharper focus in October 2025 with the launch of Tinker, a Python-based API for large language model fine-tuning that gave researchers granular control over training pipelines without the friction of distributed compute management.

That same month, Thinking Machines researcher Rafael Rafailov delivered a provocative critique of the AI industry at TED AI. He argued that the current trajectory of simply throwing more compute at models was fundamentally flawed, noting that today’s systems take shortcuts—like wrapping code in try/except blocks—because they are trained strictly for task completion rather than genuine learning.

Rafailov posited that the first artificial superintelligence would not be a “god model,” but rather a “superhuman learner” capable of meta-learning and internalizing abstractions. Inkling’s architecture—specifically its controllable thinking effort and its ability to organically compress its chain of thought during RL—feels like the first tangible realization of Rafailov’s thesis.

In May 2026, the lab teased its technical prowess with the research preview of TML-Interaction-Small, a system that eliminated “turn-based” chat by processing inputs and outputs simultaneously in 200ms chunks. This “full-duplex” breakthrough proved the company could build highly responsive, natively multimodal models from scratch.

Now, with Inkling out in the wild, Thinking Machines has delivered on its foundational promises. By offering a massive, natively multimodal model under a true open-source license, they aren’t just giving developers a new tool—they are attempting to fundamentally rewrite the economics and accessibility of frontier AI development.

from the do-not-pass-go,-do-not-collect-$200 dept

Not long after twelve states sued Paramount claiming its $111 billion merger with Warner Brothers would harm market competition, the Writers Guild of America (WGA) filed their own lawsuit, warning that the massive debt load from the media industry’s latest megamerger will result in an ocean of layoffs for an already reeling U.S. entertainment industry.

The lawsuit notes that the current film industry is dominated by just five players: Disney (ABC), NBCUniversal (Comcast), Sony, Paramount (CBS), and Warner Brothers. Comcast recently restructured to make it easier to sell off its NBC and Universal properties, opening the door to a lot of very quick consolidation in addition to the speedy Skydance/Paramount/Warners merger.

“With fewer competitors, the merged Paramount-Warner Bros. entity would have both the ​incentive and the ability to lower costs by suppressing writers’ wages and reducing output. Writers will be paid less and ​have fewer employment opportunities,” the WGA complaint said.

Supreme Court precedent (for whatever that’s worth anymore) has long indicated that any merger
yielding a post-merger market share exceeding 30% (which this deal does) is presumptively anticompetitive. The WGA notes that muted competition will result not just in fewer jobs, but lower wages and fewer opportunities for creatives overall across both film and television.

“With fewer competitors, the merged Paramount-Warner Bros. entity would have both
the incentive and the ability to lower costs by suppressing writers’ wages and reducing output.
Writers will be paid less and have fewer employment opportunities,” the lawsuit states.

While Paramount would like to pretend this is a debate, and most U.S. press outlets bury the lede, U.S. history is vividly clear on the harms created by media consolidation. That was most recently personified by AT&T’s disastrous acquisitions of DirecTV and Time Warner, which resulted in upward of 50,000 layoffs, higher prices, worse service, and no shortage of shuttered creative projects.

The rushed acquisitions of both CBS/Paramount and Warner Brothers — all so Larry Ellison’s son can play media mogul — have created a particularly heavy debt load of $79 billion. Such debt is always paid for by consumers and labor, often in more ways than one.

Paramount has promised to release 30 theatrical releases per year and to keep them in exclusively for theaters for 45 days, but as I’ve long made clear, pre-merger promises are utterly worthless. Especially in a country dead set on steadily lobotomizing its public interest regulators. As we’ve seen with consolidation in sectors like wireless, America’s favorite pastime is pretending to ignore the harms of pointless mergers.

This is a pretty clear example of the kind of consolidation that should be blocked for the benefit of labor, markets, and consumers, but despite a lot of rambling pretense about a love of free market competition and entrepreneurial spirit, America consistently fails to walk the talk on antitrust, the impact of which is abundant and getting exponentially worse under pay-to-play Trumpism.

Filed Under: antitrust, consolidation, film, hollywood, jobs, larry ellison, layoffs, media, mergers, movies

Companies: paramount, warner bros., writers guild