SMRTR AI

AI stories

Get hand-picked daily summaries of the best, most informative AI articles from around the web. Browse recent summaries, then jump to the original source when you want the full article.

Latest story• Aug 12, 2025

Anthropic’s Claude AI model can now handle longer prompts

Anthropic has expanded Claude Sonnet 4's context window to 1 million tokens, allowing enterprise API customers to process inputs equivalent to 750,000 words or 75,000 lines of code. This fivefold increase exceeds OpenAI's GPT-5 capacity and aims to strengthen Anthropic's position with AI coding platforms like GitHub Copilot and Cursor, particularly as they face growing competition from OpenAI's competitively priced...

Read summary Original source

Latest stories

TextQuests: How Good are LLMs at Text-Based Video Games?

Researchers have introduced TextQuests, a new benchmark testing large language models (LLMs) on 25 classic text-based video games. This evaluation measures how well AI agents can reason over long contexts and learn through...

Read summary Original

Why the best lawyers are learning to talk to AI

Attorneys who adapt to AI by guiding it through prompts for faster drafting and research, while applying their judgment, will remain relevant and valuable in a legal landscape increasingly influenced by this technology.

Read summary Original

Is the A.I. Boom Turning Into an A.I. Bubble?

The AI boom is showing concerning parallels to the dot-com bubble of the late 1990s, with companies like Nvidia reaching record valuations and Palantir trading at 600 times its earnings. While AI companies enjoy unprecedented...

Read summary Original

Amazon's New Alexa AI Sounds Like a Dystopian Nightmare

Amazon's Alexa+ update combines generative AI with increased advertising, frustrating tech journalists who tested it. The new system constantly pushes product purchases while breaking basic functions like alarms and news...

Read summary Original

Evaluating LLMs Playing Text Adventures

An evaluation of language models in text adventure games revealed significant performance differences. Using an achievement-based scoring system, researchers tested models across seven games. Gemini 2.5 Flash emerged as the most...

Read summary Original

SMRTR AI

Get the next batch of curated stories in your inbox.

This archive is built from SMRTR newsletter stories. Subscribe for hand-picked stories without the extra noise.

More recent stories

An AI Model for the Brain Is Coming to the ICU

A groundbreaking collaboration between Cleveland Clinic and San Francisco startup Piramidal aims to revolutionize intensive care monitoring through artificial intelligence that "reads" brainwaves in real time. Currently,...

Read summary Original

AI Industry Warns That New Lawsuit Could Destroy It Entirely

Anthropic faces a potentially devastating class-action lawsuit representing millions of writers whose books were allegedly pirated for AI training. A federal judge ruled that all affected authors can join the suit, exposing the...

Read summary Original

Claude just learned a useful ChatGPT trick

Claude's memory feature, now available for Enterprise, Team, and Max subscribers, allows referencing past conversations, with users able to toggle this function in settings and broader access planned for other tiers soon.

Read summary Original

4 things we learned from OpenAI’s GPT-5 Reddit AMA

OpenAI's Reddit AMA revealed plans to reinstate GPT-4 for Plus users, address content filtering concerns, introduce a new subscription tier, and improve GPT-5's performance while increasing rate limits for Plus subscribers.

Read summary Original

How you can still access GPT-4o, o3, and other older models in ChatGPT

After backlash, OpenAI has restored access to legacy models like GPT-4o for paid users. ChatGPT Plus subscribers ($20/month) can now toggle on "Show legacy models" in settings to access GPT-4o, while Pro users ($200/month) get...

Read summary Original

Sam Altman says the super-powerful ChatGPT-5 Pro might be coming to Plus accounts, but with one big limitation

OpenAI CEO Sam Altman hinted that ChatGPT Plus subscribers might soon get limited access to the ChatGPT-5 Pro model, currently exclusive to $200/month Pro users. Plus subscribers would receive a few GPT-5 Pro queries monthly,...

Read summary Original

Fine-tune OpenAI GPT-OSS models on Amazon SageMaker AI using Hugging Face libraries

OpenAI's GPT-OSS models are now available on AWS through SageMaker AI and Bedrock. These Mixture-of-Experts models activate a subset of parameters per token, offering high reasoning performance with reduced compute costs. They...

Read summary Original

Here are all the GPT-5 updates OpenAI has rolled out since launch

OpenAI's GPT-5 launch faced user disappointment, with complaints that it underperformed GPT-4. In response, OpenAI increased message rate limits for Plus subscribers, restored access to legacy models, improved model...

Read summary Original

Deepfake detectors are coming of age, at a time of dire need

Deepfake detection technology is advancing but still imperfect, with current systems achieving about 90% accuracy as fraud threats grow. Deloitte estimates deepfake fraud will cost the US up to $40 billion by 2027, while experts...

Read summary Original

The hidden economics of AI: balancing innovation with reality

Businesses struggle to scale AI solutions despite significant investment, with only 36% succeeding and 13% achieving enterprise impact, due to leaders overlooking implementation costs, regulatory issues, and sustainability...

Read summary Original

LLMs’ “simulated reasoning” abilities are a “brittle mirage,” researchers find

Chain-of-thought AI models perform well on familiar problems but fail dramatically when faced with new scenarios that require logical generalization. University of Arizona researchers found that these models aren't true reasoners...

Read summary Original

GPT-5 Is Turning Into a Disaster

The much-anticipated GPT-5 launch has flopped, with users demanding the return of previous versions. OpenAI quickly reinstated GPT-4o for paid subscribers following backlash. The new model suffers from inconsistent performance,...

Read summary Original

Tooling ≠ Glue: Why changing AI workflows still feels like duct tape

Despite AI development tool advances, swapping components in AI workflows remains difficult. What should be modular systems are brittle patchworks where changing one element often breaks everything. This occurs due to leaky...

Read summary Original

Vibe Coding: How SMBs Can Build AI Apps Using Plain Language

Vibe coding allows small businesses to create AI applications using plain language instead of code. This approach, coined by OpenAI's Andrej Karpathy, enables non-technical users to build prototypes quickly by describing their...

Read summary Original

AI Designs Absurd "Super Safe Sub" for Billionaires to Ride Into the Depths of the Ocean

A mysterious "emergency baby" and a "proutic haripotator" are just some of the nonsensical features included in an AI-designed submarine that supposedly outperforms the ill-fated Titan submersible. When Bluesky user Keith Ng...

Read summary Original

POML: Prompt Orchestration Markup Language

POML (Prompt Orchestration Markup Language) brings HTML-like structure to prompt engineering for Large Language Models, addressing common development challenges. The language features structured markup with semantic components,...

Read summary Original

Getting Closer To AI Internet Protocols

MIT's NANDA protocol enables AI agents to connect and interact online, using "interaction coupons" for agent discovery, trust-building, and value exchange across networked AI systems.

Read summary Original

OpenAI priced GPT-5 so low, it may spark a price war

OpenAI's newly launched GPT-5 model offers competitive pricing that may spark an industry-wide price war. Despite CEO Sam Altman's bold claims, performance tests show it only marginally outperforms or slightly lags behind...

Read summary Original

Back to all archives Load more stories