Wiredframe Radar logo

Category intelligence

Social Media Briefing — February 18, 2026

449 current items analyzed and ranked.

Executive synthesis

Social Media Summary

Anthropic's Claude Sonnet 4.6 launch dominated the day. Mike Krieger (Anthropic CPO) announced it approaches Opus-level intelligence at Sonnet pricing, with computer use as the standout capability. Boris Cherny confirmed it's live in Claude Code, and Perplexity rolled it out immediately to Pro/Max subscribers.

  • The Figma + Claude Code integration generated massive excitement (800K+ views), enabling AI-generated UI to flow directly into Figma as editable design frames—a concrete milestone for AI-assisted design workflows
  • Andriy Burkov went viral with a novel finding: simply repeating a prompt twice significantly boosts LLM accuracy across benchmarks, a zero-cost technique that surprised the community
  • John Carmack proposed a GPU preemption mechanism for research clusters, offering deep technical insight into infrastructure pain points
  • Greg Brockman posted a detailed hiring call for OpenAI infrastructure and security engineers, offering rare insider perspective on how AI is reshaping internal engineering
  • Hugo Larochelle shared independent research replicating METR's exponential trend in LLM agent time horizons, while Nathan Lambert analyzed how RL research in academia is shifting from benchmark-maxing toward more robust methodologies in 2026
  • Lambert also offered a contrarian take: open models are doing better than expected but show no evidence of closing the frontier gap permanently

Key Themes

Claude Sonnet 4.6 Launch · 12Figma + Claude Code Integration · 5AI Agents: Infrastructure, Security & Capabilities · 8GPU Infrastructure & Scheduling · 1Prompt Engineering & LLM Optimization · 3RL Research Evolution in 2026 · 2Frontier Model Comparisons · 2Open vs Closed Models · 3Anthropic: Partnerships, Ethics & Constitution · 6Agent Platforms & Personal Agents · 8

Primary evidence

Top Ranked Signals

Social Twitter Feb 17

Sonnet 4.6 is now live in Claude Code. It's cheaper than Opus 4.6 and nears Opus-level intelligence,...

By @bcherny

88 score
AI Analysis

Boris Cherny (Anthropic) announces Sonnet 4.6 is live in Claude Code: cheaper than Opus 4.6, nears Opus-level intelligence, devs in early testing often preferred it to Opus 4.5. Now default for Pro and Team plans.

Sonnet 4.6 is now live in Claude Code. It's cheaper than Opus 4.6 and nears Opus-level intelligence, and devs in early testing often preferred it to Opus 4.5. Now the default for Pro and Team plans.
Claude Sonnet 4.6 LaunchClaude CodeModel ReleaseCoding AI
Social Twitter Feb 17

Figma just shipped the ability to bring UI work done in Claude Code straight into Figma as editable ...

By @trq212

88 score
AI Analysis

trq212 announces Figma has shipped the ability to bring UI work done in Claude Code straight into Figma as editable design frames, enabling idea exploration, multi-page flow viewing, and UX reimagination.

Figma just shipped the ability to bring UI work done in Claude Code straight into Figma as editable design frames. Use this to explore new ideas in Figma, view multi-page flows on the canvas, or reimagine user experiences. t.co/OwBbfRpvch
figma-integrationclaude-codedesign-to-codeproduct-announcementdeveloper-tools
Social Twitter Feb 17

If you’re an infrastructure or security engineer, now is the best time to join OpenAI. It’s hard no...

By @gdb

82 score
AI Analysis

Greg Brockman (OpenAI co-founder) posts a detailed hiring call for infrastructure and security engineers at OpenAI, describing how AI is transforming engineering and what skills matter going forward.

If you’re an infrastructure or security engineer, now is the best time to join OpenAI. It’s hard not to be inspired by what today’s coding tools are capable of, and we have line of sight to making them much better. While our core ML infrastructure problems remain much the same as always — training and inferencing models at scale, co-designing end-to-end for maximum effect, managing complexity and maintaining fast iteration — what it feels like to solve these problems in practice is changing fas
OpenAI strategyAI engineeringAI agents infrastructurefuture of software engineeringhiring
Social Twitter Feb 17

LLMs process text from left to right — each token can only look back at what came before it, never f...

By @burkov

82 score
AI Analysis

Burkov explains a research finding that simply sending a prompt twice in a row significantly improves LLM accuracy across multiple benchmarks and models (Gemini, ChatGPT, Claude, DeepSeek), with no additional output length or meaningful latency increase. One model improved from 21% to 97% on a task.

LLMs process text from left to right — each token can only look back at what came before it, never forward. This means that when you write a long prompt with context at the beginning and a question at the end, the model answers the question having "seen" the context, but the context tokens were generated without any awareness of what question was coming. This asymmetry is a basic structural property of how these models work. The paper asks what happens if you just send the prompt twice in a row
prompt engineeringLLM optimizationresearch breakthroughpractical AI techniques
Social Twitter Feb 17

The glory work of GPU scheduling is in the frontier data centers with hundreds of thousands of GPUs,...

By @ID_AA_Carmack

82 score
AI Analysis

John Carmack discusses GPU scheduling challenges for research clusters, proposing a transparent preemption mechanism using Nvidia UVM and MPS to pause low-priority GPU tasks for higher-priority ones, enabling task switching at tens-of-seconds granularity rather than milliseconds.

The glory work of GPU scheduling is in the frontier data centers with hundreds of thousands of GPUs, but a lot of research work is done with single GPU jobs on modest clusters, and the scheduling leaves much to be desired. I wish there were a clean way to preempt GPU tasks, so long running tasks could be transparently paused to allow higher priority tasks to get the minimum time-to-results. Manual checkpointing and cooperative multitasking is an option, but it complicates codebases and is ferti
GPU schedulingAI infrastructureresearch computingNvidiasystems engineering
Social Twitter Feb 17

Claude Code + Figma = 👌

By @bcherny

78 score
AI Analysis

Boris Cherny showcases Claude Code + Figma integration with extremely high engagement (465K views, 2.9K likes).

Claude Code + Figma = 👌
Claude CodeDesign-to-CodeAI Development ToolsFigma Integration
Social Twitter Feb 17

Computer use is the standout. For coding, it’s less prone to overengineering than Opus 4.5 and more ...

By @mikeyk

78 score
AI Analysis

Mike Krieger (Anthropic CPO) highlights Sonnet 4.6's standout features: computer use capability, less prone to overengineering than Opus 4.5 for coding, more consistent over long sessions, and 1M context window in beta on API.

Computer use is the standout. For coding, it’s less prone to overengineering than Opus 4.5 and more consistent over long sessions. And 1M context window in beta on the API. We can't wait to see what you build!
claude-sonnet-4.6anthropicmodel-releasecomputer-usecontext-windowcoding-ai
Social Twitter Feb 17

Ever since @METR_Evals' fascinating work on LLM agent time horizons, I've wanted to see other attemp...

By @hugo_larochelle

75 score
AI Analysis

Hugo Larochelle shares follow-up research on LLM agent time horizons, referencing METR's work. Their independent analysis also finds an exponential trend but estimates shorter horizons for recent models (~2h vs METR's ~5h).

Ever since @METR_Evals' fascinating work on LLM agent time horizons, I've wanted to see other attempts to draw conclusions from the same data. In a separate approach by Fengyuan and Jay, we too infer an exponential, but with shorter horizons for recent models (~2h vs METR's ~5h)
AI agentsAI capabilities forecastingbenchmarksresearch
Social Twitter Feb 17

RL research in academia is shaping up to be much more exciting in 2026 than the last half of 2025. H...

By @natolambert

75 score
AI Analysis

Nathan Lambert provides an insightful analysis of how RL research in academia is evolving in 2026 - shifting from benchmark-maxing to more robust research focused on tool-use algorithms, procedural environment generation, and generalization across domains.

RL research in academia is shaping up to be much more exciting in 2026 than the last half of 2025. Here's why I see it as a healthier transition from a bit of benchmaxing to more interesting and robust research (dare I say RL generalizing across many domains?) 2025 was largely a setup year, where due to the simplicity of the environments, insignificant algorithmic changes could appear more valuable than they were. Today, substantial work is focusing on algorithms for tool-use, procedural envir
rl-researchai-researchgeneralizationacademic-aienvironment-generationai-industry-analysis
Social Twitter Feb 17

We've signed an MOU with the Government of Rwanda—the first partnership of its kind in Africa—to bri...

By @AnthropicAI

73 score
AI Analysis

Anthropic announces an MOU with the Government of Rwanda - their first partnership in Africa - to bring AI to health, education, and public sectors.

We've signed an MOU with the Government of Rwanda—the first partnership of its kind in Africa—to bring AI to health, education, and other public sectors. Read more: t.co/txgEScvKtP
AI policyAnthropicAfrica AIAI for developmentpublic sector AI
Social Twitter Feb 17

It really isn’t that hard to see the jagged frontier of AI. Just think about the parts of your job t...

By @emollick

72 score
AI Analysis

Ethan Mollick discusses the 'jagged frontier' of AI - encouraging people to identify parts of their jobs that AI cannot do even if agents improve 10x, and notes that heavy AI usage improves these assessments.

It really isn’t that hard to see the jagged frontier of AI. Just think about the parts of your job that are vital but that you would be insane to expect an AI to do, even if agents get 10x better. Thats the frontier The more you use AI the more accurate those assessments will be
AI capabilities and limitationsAI adoption strategyfuture of work