Wiredframe Radar logo

Category intelligence

Social Media Briefing — January 3, 2026

393 current items analyzed and ranked.

Executive synthesis

Social Media Summary

Boris Cherny, creator of Claude Code, dominated AI discussions with a comprehensive 7.2M-view thread detailing his personal workflows—running 5+ parallel Claude instances, using shared CLAUDE.md files across teams, and emphasizing verification as the key to 2-3x quality improvement.

  • Nathan Lambert released major updates to his RLHF book (now 200 pages) and shared a curated timeline of 26 reasoning model technical reports from DeepSeek to others
  • Greg Brockman sparked debate claiming Rust is ideal for AI agents due to compiler guarantees; received 1.4M views
  • Erik Brynjolfsson (MIT) predicted the rise of 'Chief Question Officers' who manage AI agent fleets
  • Andriy Burkov made bold claims that Anthropic has unmatched AI know-how and Claude Code has made junior developers obsolete for greenfield projects

Key Themes

Claude Code Workflow · 15Reasoning Models & RLHF · 6Claude Code & AI Coding Agents · 28Parallel Agent Execution · 6AI Verification & Quality · 5OpenAI Codex & Agent Development · 3Team AI Collaboration · 6AI Agents & Agentic Systems · 12Workflow Automation · 5Developer Productivity & Future of Work · 5

Primary evidence

Top Ranked Signals

Social Twitter Jan 2

I'm Boris and I created Claude Code. Lots of people have asked how I use Claude Code, so I wanted to...

By @bcherny

98 score
AI Analysis
Boris Cherny, creator of Claude Code, introduces a comprehensive thread sharing his personal setup and workflow for using Claude Code. Notes that his setup is 'surprisingly vanilla' and emphasizes the tool works great out of the box with customization being optional.
I'm Boris and I created Claude Code. Lots of people have asked how I use Claude Code, so I wanted to show off my setup a bit. My setup might be surprisingly vanilla! Claude Code works great out of the box, so I personally don't customize it much. There is no one correct way to use Claude Code: we intentionally build it in a way that you can use it, customize it, and hack it however you like. Each person on the Claude Code team uses it very differently. So, here goes.
Claude Code WorkflowAI-Assisted DevelopmentDeveloper Tooling
Social Twitter Jan 2

RLHF Book status update: lot's of great changes. Over the past month I've been doing a top to botto...

By @natolambert

95 score
AI Analysis
Major RLHF Book update: expanded from 150 to 200 pages with new RL algorithms (GSPO, CISPO), reasoning model coverage, rubrics for RLVR, and fixes to original RLHF diagram from 2022
RLHF Book status update: lot's of great changes. Over the past month I've been doing a top to bottom update to the RLHF book. All of these changes are reflected on the website rlhfbook dot com, and will soon be translated to the Manning early access version (MEAP), and then more improvements for the physical copy. Overall, this took the PDF from ~150 to ~200 pages, the book is much more well rounded now. Some of the larger changes:
  • Updates to the RL chapter to add more algorithms like GSPO
rlhfreinforcement-learningreasoning-modelsai-educationtechnical-content
Social Twitter Jan 2

13/ A final tip: probably the most important thing to get great results out of Claude Code -- give C...

By @bcherny

93 score
AI Analysis
Key tip: Give Claude a way to verify its work for 2-3x quality improvement. Claude tests every change to claude.ai using Chrome extension - opens browser, tests UI, iterates until code works and UX is good.
13/ A final tip: probably the most important thing to get great results out of Claude Code -- give Claude a way to verify its work. If Claude has that feedback loop, it will 2-3x the quality of the final result. Claude tests every single change I land to t.co/pEWPQoSq5t using the Claude Chrome extension. It opens a browser, tests the UI, and iterates until the code works and the UX feels good. Verification looks different for each domain. It might be as simple as running a bash command
AI VerificationQuality AssuranceFeedback Loops
Social Twitter Jan 2

1/ I run 5 Claudes in parallel in my terminal. I number my tabs 1-5, and use system notifications to...

By @bcherny

92 score
AI Analysis
Boris runs 5 Claude instances in parallel in terminal tabs numbered 1-5, using system notifications to know when Claude needs input.
1/ I run 5 Claudes in parallel in my terminal. I number my tabs 1-5, and use system notifications to know when a Claude needs input t.co/nmRJ5km3oZ t.co/CJaX1rUgiH
Parallel Agent ExecutionDeveloper ProductivityClaude Code Workflow
Social Twitter Jan 2

Reasoning model reports I recommend reading: 2025-01-22 - DeepSeek R1 - https://t.co/wUGy9WMOSE 202...

By @natolambert

92 score
AI Analysis
Comprehensive curated list of 26 reasoning model technical reports from DeepSeek R1 through DeepSeek V3.2, spanning Jan-Dec 2025
Reasoning model reports I recommend reading: 2025-01-22 - DeepSeek R1 - t.co/wUGy9WMOSE 2025-01-22 - Kimi 1.5 - t.co/4ZlY5VsOJY 2025-03-31 - Open-Reasoner-Zero - t.co/kqTw6RNQs9 2025-04-10 - Seed-Thinking 1.5 - t.co/hinZOjQzWq 2025-04-30 - Phi-4 Reasoning - t.co/zJIDMeYOV0 2025-05-02 - Llama-Nemotron - t.co/wEJ0WMTsFv 2025-05-12 - INTELLECT-2 - t.co/akoi15sdBy 2025-05-12 - Xiaomi MiMo - t.co/Cf3bT3ijzv 2025-05-14 - Qwen 3 - https:/
reasoning-modelsllm-researchtechnical-reportsopen-source-ai
Social Twitter Jan 2

2/ I also run 5-10 Claudes on https://t.co/pEWPQoSq5t, in parallel with my local Claudes. As I code ...

By @bcherny

91 score
AI Analysis
Boris also runs 5-10 Claudes on claude.ai in parallel with local instances, using teleport feature to move sessions between CLI/web/phone. Starts sessions from iOS app throughout the day.
2/ I also run 5-10 Claudes on t.co/pEWPQoSq5t, in parallel with my local Claudes. As I code in my terminal, I will often hand off local sessions to web (using &), or manually kick off sessions in Chrome, and sometimes I will --teleport back and forth. I also start a few sessions from my phone (from the Claude iOS app) every morning and throughout the day, and check in on them later.
Parallel Agent ExecutionMulti-Platform AI WorkflowClaude Code Workflow
Social Twitter Jan 2

3/ I use Opus 4.5 with thinking for everything. It's the best coding model I've ever used, and even ...

By @bcherny

90 score
AI Analysis
Boris recommends Opus 4.5 with thinking for everything, calling it the best coding model. Despite being larger/slower than Sonnet, reduced steering needs and better tool use make it faster overall.
3/ I use Opus 4.5 with thinking for everything. It's the best coding model I've ever used, and even though it's bigger & slower than Sonnet, since you have to steer it less and it's better at tool use, it is almost always faster than using a smaller model in the end.
AI Model SelectionOpus 4.5Coding Model Performance
Social Twitter Jan 2

4/ Our team shares a single https://t.co/pp5TJkWmFE for the Claude Code repo. We check it into git, ...

By @bcherny

89 score
AI Analysis
Continuing our coverage from yesterday, Claude Code team shares a single CLAUDE.md file checked into git. Entire team contributes updates multiple times weekly, adding corrections whenever Claude makes mistakes to prevent recurrence.
4/ Our team shares a single t.co/pp5TJkWmFE for the Claude Code repo. We check it into git, and the whole team contributes multiple times a week. Anytime we see Claude do something incorrectly we add it to the t.co/pp5TJkWmFE, so Claude knows not to do it next time. Other teams maintain their own t.co/pp5TJkWmFE's. It is each team's job to keep theirs up to date.
Team AI CollaborationKnowledge ManagementCLAUDE.md Best Practices
Social Twitter Jan 2

The only AI company that has any significant know-how today is Anthropic. No one is even close to wh...

By @burkov

88 score
AI Analysis
Burkov claims Anthropic is the only AI company with significant know-how, stating Claude Code has made junior/mid-level developers obsolete for projects now buildable in hours instead of months
The only AI company that has any significant know-how today is Anthropic. No one is even close to what Claude Code can do with code. This is the only use case that has a measurable impact on society: The concept of a junior or mid-level developer is gone, and the projects that these two categories of developers were capable of building in months are now built in hours or, in some cases, days. All other AI use cases are either fun but useless, or not ready for prime time, or niche, or it's impo
Claude CodeAnthropicDeveloper DisplacementIndustry AnalysisAI Coding
Social Twitter Jan 2

5/ During code review, I will often tag @.claude on my coworkers' PRs to add something to the https:...

By @bcherny

87 score
AI Analysis
During code review, Boris tags @.claude on coworkers' PRs to add learnings to CLAUDE.md as part of the PR using GitHub Action. Calls this 'Compounding Engineering'.
5/ During code review, I will often tag @.claude on my coworkers' PRs to add something to the t.co/v4FOLUBHz9 as part of the PR. We use the Claude Code Github action (/install-github-action) for this. It's our version of @danshipper's Compounding Engineering t.co/VIQYZ2hFq5
Code Review AutomationCompounding EngineeringGitHub Integration
Social Twitter Jan 2

6/ Most sessions start in Plan mode (shift+tab twice). If my goal is to write a Pull Request, I will...

By @bcherny

86 score
AI Analysis
Boris starts most sessions in Plan mode (shift+tab twice). For PRs, he iterates on the plan with Claude before switching to auto-accept mode, emphasizing that 'a good plan is really important'.
6/ Most sessions start in Plan mode (shift+tab twice). If my goal is to write a Pull Request, I will use Plan mode, and go back and forth with Claude until I like its plan. From there, I switch into auto-accept edits mode and Claude can usually 1-shot it. A good plan is really important!
AI Planning StrategyClaude Code WorkflowDeveloper Methodology