Wiredframe Radar logo

Category intelligence

Social Media Briefing — January 25, 2026

502 current items analyzed and ranked.

Executive synthesis

Social Media Summary

The AI community debated the future of software engineering and AGI timelines with sharp contrasts. Yann LeCun pushed back firmly on AGI hype, arguing that superhuman task performance has repeatedly been mistaken for human-level intelligence. Greg Brockman framed the paradigm shift toward agent-first development, highlighting Cursor + GPT-5.2 autonomously building a browser.

  • Santiago Pino drew viral attention (364K views) to the contradiction of Anthropic's CEO declaring software engineering dead while the company continues hiring engineers
  • Ethan Mollick and swyx both noted Claude's lead over competitors in spreadsheet integration, with swyx claiming Anthropic is 0.5-3 years ahead of Gemini
  • Levelsio reported testing vanilla Claude for autonomous product building—found it's not quite there yet
  • GPT-5.2 Pro demonstrated capability by identifying a flaw in its own benchmark math problems
  • Erik Bernhardsson provided unique infrastructure insight: Blackwell GPUs remain underused due to unoptimized kernels, driving Hopper price increases

Key Themes

GPT-5.2 Capabilities · 7AI vs Software Engineering Future · 5AI Agents & Autonomous Coding · 6Claude Excel/Spreadsheet Superiority · 3AGI Skepticism & Timelines · 4Gemini UX & Technical Issues · 5Claude vs Microsoft AI Tools · 4Embodied AI & Humanoid Robotics · 8AI Agent Architecture & Strategy · 8AI Capabilities Assessment · 2

Primary evidence

Top Ranked Signals

Social Twitter Jan 24

@redmonduser @RichardSSutton You are a victim of the same delusion as numerous folks who have believ...

By @ylecun

92 score
AI Analysis

YLecun argues that superhuman AI performance on specific tasks (code, math, Go, chess, etc.) has repeatedly been mistaken as harbinger of human-level AI throughout history

@redmonduser @RichardSSutton You are a victim of the same delusion as numerous folks who have believed in past decades that superhuman performance by computers in one task was a harbinger of human-level AI. It happened with code generation, math, chatbots, go players, robot acrobats, Jeopardy-playing systems, cars driving themselves in the desert, chess players, inference engines, checker players, compilers, equation solvers....
agi-skepticismai-capabilitiesai-hype-cycle
Social Twitter Jan 24

inspiring how agent-first software engineering raises both the floor (much easier for anyone to buil...

By @gdb

88 score
AI Analysis

Greg Brockman observes that agent-first software engineering raises both floor and ceiling of what people can build - easier for beginners, more powerful for experts

inspiring how agent-first software engineering raises both the floor (much easier for anyone to build) and the ceiling (experts can build so much more) of what people can create
ai-agentssoftware-engineeringai-democratization
Social Twitter Jan 24

Can someone explain to me why Anthropic's CEO keeps saying Software Engineering is dead, yet his com...

By @svpino

88 score
AI Analysis

Santiago Pino questions why Anthropic CEO Dario Amodei keeps saying software engineering is dead while Anthropic continues to hire software engineers - highlighting a disconnect between AI hype and actual industry practices

Can someone explain to me why Anthropic's CEO keeps saying Software Engineering is dead, yet his company is still hiring Software Engineers?
AI industry criticismSoftware engineering futureAI hype vs reality
Social Twitter Jan 24

Claude in Excel is really good. Its weird that using Microsoft's own Excel agent using Claude 4.5 o...

By @emollick

85 score
AI Analysis

Emollick finds Claude in Excel superior to Microsoft's own Excel agent using Claude 4.5, because Claude does own analysis while Microsoft agent relies on VLOOKUPs

Claude in Excel is really good. Its weird that using Microsoft's own Excel agent using Claude 4.5 often yields weaker answers, It seems to be because the Excel agent relies on Excel alone (VLOOKUPs, etc) while Claude in Excel does its own analysis and uses Excel for output.
claude-toolsmicrosoft-copilotai-productivitymodel-comparison
Social Twitter Jan 24

16M impressions in 24 hours. if you’ve ever tried Claude in Sheets or Claude in Excel you will know ...

By @swyx

82 score
AI Analysis

swyx compares Claude in Sheets/Excel vs Gemini, claiming Anthropic is 0.5-3 years ahead on spreadsheet integration. Notes 16M impressions in 24 hours on the topic, suggesting major interest in AI-powered spreadsheet tools.

16M impressions in 24 hours. if you’ve ever tried Claude in Sheets or Claude in Excel you will know how much more intelligent it is compared to Gemini in Sheets i have two current measures of Google-GDM product integration right now:
  • how long does it take Google to put a non nerfed Gemini Pro into Sheets
  • how long does it take to make Gemini 3.5 halfway decent at Sheets manipulation and formula
clock has started, Anthropic is between 0.5 to 3 years ahead on this one
claude_capabilitiesgoogle_vs_anthropicproductivity_tools
Social Twitter Jan 24

@shaneguML Hi Shane, I hope you are well. This partnership with Google is especially meaningful to...

By @hardmaru

78 score
AI Analysis

Hardmaru (Sakana AI) confirms partnership with Google DeepMind, tags Demis Hassabis and Jeff Dean

@shaneguML Hi Shane, I hope you are well. This partnership with Google is especially meaningful to me and for Sakana AI. Just to clarify, @GoogleDeepMind is indeed involved in our collaboration. We look forward to working with the DeepMind team: @demishassabis @JeffDean @heiga_zen David
ai-partnershipssakana-aigoogle-deepmind
Social Twitter Jan 24

So, Claude told me to take a walk for 30 minutes, so I tried Gemini Pro for the first time in a coup...

By @burkov

78 score
AI Analysis

Burkov expresses frustration with Gemini Pro's hedging language ('typically', 'might', 'should'), calling it dangerous for mental health. Notes Claude told him to take a walk, so he tried Gemini instead.

So, Claude told me to take a walk for 30 minutes, so I tried Gemini Pro for the first time in a couple of months. Started to feel angry after about 10 minutes of using it. My triggers:
  • "This error is *typically* due to..., so here's the fix."
  • "This *might* be the reason for this issue, so here's the fix."
  • "Change A to B and this *should* fix it."
Working like this is plain dangerous for your mental health. I went for a walk.
llm_ux_issuesgemini_critiquedeveloper_experience
Social Twitter Jan 24

The difference between the software being the point (Microsoft's perspective) and the output being t...

By @emollick

75 score
AI Analysis

Emollick contrasts Microsoft's 'software is the point' vs Anthropic's 'output is the point' philosophy for agentic workspace competition

The difference between the software being the point (Microsoft's perspective) and the output being the point (Anthropic's perspective). Do you care about the tool, or about the answer? Seems like that is going be a very consequential decision for who wins the agentic work space.
ai-strategymicrosoft-vs-anthropicagentic-ai
Social Twitter Jan 24

They will keep killing software engineering every 6 months until morale improves.

By @svpino

75 score
AI Analysis

svpino sarcastically comments 'They will keep killing software engineering every 6 months until morale improves' - satirizing recurring AI hype cycles about replacing developers.

They will keep killing software engineering every 6 months until morale improves.
ai_hype_fatiguesoftware_engineering_futureindustry_sentiment
Social Twitter Jan 24

I actually tried this and we're not there with vanilla Claude yet, we might just need to add some pl...

By @levelsio

75 score
AI Analysis

Continuing his exploration from Social yesterday, Levelsio reports testing vanilla Claude to autonomously build products and register domains - found it's not quite there yet but suggests plugins for domain registration could close the gap

I actually tried this and we're not there with vanilla Claude yet, we might just need to add some plugins so it can register domain names? I don't see the issue t.co/rBaRYVzpww
AI capabilities assessmentClaude limitationsAutonomous AI agents