Wiredframe Radar logo

Category intelligence

Social Media Briefing — May 14, 2026

443 current items analyzed and ranked.

Executive synthesis

Social Media Summary

The AI community focused on two dominant stories: OpenAI's aggressive Codex pricing play and the UK AI Security Institute's alarming findings on frontier model cyber capabilities.

Research and open-source developments also drew attention: an EMNLP 2025 paper showed competitive reasoning models can be trained in just 26 minutes with 1,000 examples, while Hugging Face open-sourced a CLI agent automating full ML research loops. Clement Delangue urged the American AI community to support open international collaboration during the Trump-Xi summit.

Key Themes

AI Safety & Cybersecurity Capabilities · 2Frontier Model Safety & Regulation · 4UK AI Safety Evaluation of Frontier Models · 3Codex & AI Coding Competition · 6Enterprise AI Adoption & Competition · 7Anthropic Competitive Strategy & AI Governance · 2AI Agent Infrastructure & Security · 10Open Source Models Catching Up · 3Enterprise AI Transformation & Governance · 8Training Efficiency & Research · 5

Primary evidence

Top Ranked Signals

Social Twitter May 13

codex is the best AI coding product and we want to make it easy to try. for the next 30 days, we ar...

By @sama

92 score
AI Analysis

Sam Altman announces Codex as best AI coding product, offers companies 2 months free usage for next 30 days to encourage switching

codex is the best AI coding product and we want to make it easy to try. for the next 30 days, we are giving companies that want to try switching over two months of free codex usage.
codex_enterprise_adoptionai_coding_toolscompetitive_strategy
Social Twitter May 13

The UK AISI found Mythos Preview is the first model to solve both their cyber ranges end-to-end. No ...

By @bcherny

92 score
AI Analysis

Boris Cherny reports that UK AISI found Claude Mythos Preview is the first model to solve both their cyber ranges end-to-end, including the previously unsolved 'Cooling Tower' range. Mentions Glasswing work.

The UK AISI found Mythos Preview is the first model to solve both their cyber ranges end-to-end. No model had ever solved the AISI’s “Cooling Tower” cyber range before. We're getting it to defenders as fast as we responsibly can. More to come on our Glasswing work soon.
ai_safetycybersecuritymodel_capabilitiesanthropicbenchmarks
Social Twitter May 13

The UK’s state AI Security iIstitute findings: 1) Mythos is a big gain in cyber capabilities. But so...

By @emollick

88 score
AI Analysis

Emollick summarizes UK AI Security Institute findings: Mythos is major cyber capability gain (as is GPT-5.5), hard to establish upper bounds on these models, capability doubling time is 4.5 months

The UK’s state AI Security iIstitute findings: 1) Mythos is a big gain in cyber capabilities. But so is GPT-5.5 2) It is hard to establish an upper bound on Mythos/GPT-5.5, which appear to be limited by tokens used, rather than ability. 3) Capability doubling time is 4.5 months t.co/k9bPGiAUwt
ai_safetyfrontier_model_capabilitiescyber_securitygovernment_evaluationcapability_growth
Social Bluesky May 13

The UK’s state AI Security iIstitute findings on latest AI models: 1) Mythos is a big gain in cyber ...

By @emollick.bsky.social

88 score
AI Analysis

Ethan Mollick summarizes UK AI Safety Institute findings: Mythos shows big cyber capability gains alongside GPT-5.5, both models appear token-limited rather than ability-limited, and AI capability doubling time is 4.5 months

The UK’s state AI Security iIstitute findings on latest AI models: 1) Mythos is a big gain in cyber capabilities. But so is GPT-5.5 2) It is hard to establish an upper bound on Mythos/GPT-5.5, which appear to be limited by tokens used, rather than ability. 3) Capability doubling time is 4.5 months
ai_safetyfrontier_modelscyber_capabilitiescapability_scalinggovernment_ai_evaluationclaude_mythosgpt55
Social Twitter May 13

I don't understand the path forward for Mythos releases. Google & OpenAI will have equivalent mo...

By @emollick

82 score
AI Analysis

Emollick questions Anthropic's path forward for Mythos releases given that Google/OpenAI will have equivalent models with different guardrail approaches, asks how Anthropic escapes government approval constraints

I don't understand the path forward for Mythos releases. Google & OpenAI will have equivalent models, and they are approaching AI cyber risk guardrails differently, so they will presumably just release their versions. How does Anthropic get out of the government approval path?
ai_safetyfrontier_model_releasecompetitive_landscapegovernment_regulationanthropic_strategy
Social Twitter May 13

Big shift in enterprise AI spending: Anthropic surpassed OpenAI for the first time in April, per @tr...

By @TheRundownAI

80 score
AI Analysis

Reports that Anthropic surpassed OpenAI in enterprise AI spending for the first time in April per Ramp's AI Index. 34.4% of US businesses have paid Anthropic subs vs 32.3% for OpenAI. Anthropic quadrupled business adoption over past year while OpenAI grew 0.3%.

Big shift in enterprise AI spending: Anthropic surpassed OpenAI for the first time in April, per @tryramp's AI Index. Share of U.S. businesses with paid AI subscriptions: Anthropic: 34.4% (+3.8%) OpenAI: 32.3% (-2.9%) Over the last year, Anthropic quadrupled business adoption while OpenAI grew by just 0.3%.
enterprise_aimarket_competitionanthropicopenaibusiness_adoption
Social Twitter May 13

i get some anxiety not using the smartest-available model/settings. but sometimes i dont mind if it...

By @sama

78 score
AI Analysis

Sam Altman reflects on anxiety about not using smartest model, wonders if OpenAI should focus more on price/speed tradeoff vs price/intelligence

i get some anxiety not using the smartest-available model/settings. but sometimes i dont mind if it's really slow. i wonder if we should focus more on a price/speed tradeoff relative to a price/intelligence tradeoff.
ai_product_strategyinference_costmodel_selection
Social Twitter May 13

This EMNLP 2025 paper shows that a pretrained 32-billion-parameter model can be turned into a compet...

By @burkov

78 score
AI Analysis

Shares EMNLP 2025 paper showing a 32B parameter model can become a competitive reasoning model using only 1,000 examples and 26 minutes of training on 16 GPUs, matching o1-preview on math/science using 'budget forcing' technique with 800x fewer examples than DeepSeek R1.

This EMNLP 2025 paper shows that a pretrained 32-billion-parameter model can be turned into a competitive reasoning model using only 1,000 carefully chosen examples and 26 minutes of additional training on 16 GPUs. The authors control how much the model "thinks" before answering through a method they call budget forcing: when the model tries to end its reasoning, they suppress its end-of-thinking token and append the word "Wait," which often nudges it to reconsider and fix earlier mistakes. Wi
reasoning_modelstraining_efficiencyresearch_papers
Social Twitter May 13

The most expensive mistake in enterprise AI right now: treating FDEs as your whole transformation pl...

By @alliekmiller

78 score
AI Analysis

Following recent News coverage of AI labs launching consulting arms, Allie K Miller argues that Forward Deployed Engineers (FDEs) alone won't transform enterprises into AI-first companies. Emphasizes that tech deployment without change management, people enablement, and process reinvention will stall. Compares to Salesforce adoption challenges.

The most expensive mistake in enterprise AI right now: treating FDEs as your whole transformation plan. Forward deployed engineers (FDEs) are important for custom deployments, but they won’t fix the change management issue most enterprises are facing. It’s likely more the former that Anthropic and OpenAI will continue to prioritize (and hire into the thousands, who knows). Beyond performance and cost, it’s systems integration, ROI, and literal usefulness that drive revenue and stickiness.
enterprise AI transformationchange managementFDE modelAI adoptionAnthropicOpenAIenterprise strategy
Social Bluesky May 13

I don't understand the path forward for Mythos releases. Google & OpenAI will have equivalent models...

By @emollick.bsky.social

78 score
AI Analysis

Ethan Mollick questions Anthropic's path forward for Mythos releases given Google/OpenAI will have equivalent models but release them without going through government approval. Questions how Anthropic exits the government approval path

I don't understand the path forward for Mythos releases. Google & OpenAI will have equivalent models, and they are approaching AI cyber risk guardrails differently, so they will presumably just release their versions without as much drama. How does Anthropic get out of the government approval path?
ai_governanceanthropic_strategycompetitive_dynamicsai_safetyclaude_mythosresponsible_release
Social Twitter May 13

Hugging Face open-sourced a CLI agent that runs full research loop of an ML engineer. You give it ...

By @AlphaSignalAI

75 score
AI Analysis

Hugging Face open-sourced a CLI agent that runs full ML research loops: reads papers, walks citations, pulls datasets, writes training scripts, launches GPU jobs, and iterates. Achieved 32% on GPQA with Qwen3-1.7B (vs Claude Code's 22.99%), beat Codex on HealthBench by 60%.

Hugging Face open-sourced a CLI agent that runs full research loop of an ML engineer. You give it a prompt. It reads papers and walks citations. Then it pulls datasets and writes training scripts. It launches GPU jobs and iterates until something works. On scientific reasoning, it trained Qwen3-1.7B and pushed GPQA from 10% to 32%. Claude Code only reached 22.99%. For healthcare, it judged public datasets too weak. So it generated 1,100 synthetic samples and upsampled 50x. It bea
ai_agentsopen_sourceml_research_automationhugging_facebenchmarks
Social Twitter May 13

As President Trump meets President Xi this week, a call to the American AI community: If your start...

By @ClementDelangue

72 score
AI Analysis

Clement Delangue calls on American AI community to show support for open international AI (especially Chinese models) during Trump-Xi meeting, argues open source drives competition, jobs, and wealth

As President Trump meets President Xi this week, a call to the American AI community: If your startup, lab, non-profit or company benefits from open international AI - especially Chinese (Deepseek, Qwen, Kimi, GLM,…), please share! Open source is the most important driver of competition, jobs and wealth creation in AI today. Let’s support and promote it at critical times like this week!
open_source_aigeopoliticsus_china_relationsai_policy