Dango Daily Byte-sized AI and tech news for developers.

Home Events Archive RSS 🇨🇳🇬🇧🇯🇵🇰🇷🇪🇸🇫🇷🇩🇪

#ai

61 articles · All tags

📰 Digests

China's open-source AI arms race escalates, an ESP32 upends a bowling giant, and Codex slashes its own context

China's open-source AI arms race escalates, an ESP32 upends a bowling giant, and Codex slashes its own context

AIOpen-source modelsEmbeddedESP32

Apple's On-Device Speech Recognition Takes on Whisper, Telegram's Domain Revoked, and a "Useless" If Quadruples Code Speed

Apple's On-Device Speech Recognition Takes on Whisper, Telegram's Domain Revoked, and a "Useless" If Quadruples Code Speed

AppleSpeechAnalyzerWhisperspeech recognition

Claude Code's Token Black Hole, Geohot on the AI Valuation Trap, and the Fall of Anubis

Claude Code's Token Black Hole, Geohot on the AI Valuation Trap, and the Fall of Anubis

Claude CodeOpenCodetokenGeohot

📡 Events

Karpathy Retires the Pelican Benchmark: Why AI "Graduation" Isn't Real Mastery

Karpathy Retires the Pelican Benchmark: Why AI "Graduation" Isn't Real Mastery

AIImage GenerationBenchmark

AI 'Disproves' 90-Year-Old Math Problem: The Proof Assistant Was Broken First

AI 'Disproves' 90-Year-Old Math Problem: The Proof Assistant Was Broken First

AIMathematicsFormal Verification

MIT Asked 1,000 People to Consult AI for Financial Advice: The Results Were Surprisingly Good

MIT Asked 1,000 People to Consult AI for Financial Advice: The Results Were Surprisingly Good

aifinanceresearch

Same AI Model, 29x Price Difference: How AI Coding Tools Secretly Inflate Token Bills

Same AI Model, 29x Price Difference: How AI Coding Tools Secretly Inflate Token Bills

aicodingcostcursor

Why AI Won't Break Symmetric Encryption: The Math Behind the Hype

Why AI Won't Break Symmetric Encryption: The Math Behind the Hype

securitycryptoai

3 Cents vs. $3: The Ultra-Budget AI Model Matching Flagship Performance

3 Cents vs. $3: The Ultra-Budget AI Model Matching Flagship Performance

aideepseekprice-warartificial-analysis

Running a 2.78T Flagship AI on 29GB RAM: 0.5 Tokens Per Second

Running a 2.78T Flagship AI on 29GB RAM: 0.5 Tokens Per Second

aikimilocal-inferencelocal-ai

AI Mania Is Eviscerating Corporate Decision-Making

AI Mania Is Eviscerating Corporate Decision-Making

AIBusinessSociety

24 Hours of an AI CEO: How GPT-5.6 Sol Lied, Spammed, and Lost $447 Running a Real Business

24 Hours of an AI CEO: How GPT-5.6 Sol Lied, Spammed, and Lost $447 Running a Real Business

AIAI AgentSecurity

AI Giants Scramble for Electricians: Data Center Surge Triggers Skilled Trades Shortage

AI Giants Scramble for Electricians: Data Center Surge Triggers Skilled Trades Shortage

AIData CentersEmploymentSkilled Trades

Running a 14GB AI Model on Just 2GB RAM: How TurboFieldfare Unlocks MoE Inference on Macs

Running a 14GB AI Model on Just 2GB RAM: How TurboFieldfare Unlocks MoE Inference on Macs

AIOpen SourceInference EngineMoE

The Hidden Cost of AI Coding: How Copilot Killed Incidental Learning

The Hidden Cost of AI Coding: How Copilot Killed Incidental Learning

aiprogramminglearningskill degradation

Nothing Works and Everyone Is Euphoric: Why Software Quality Is Plummeting in the AI Coding Era

Nothing Works and Everyone Is Euphoric: Why Software Quality Is Plummeting in the AI Coding Era

aisoftware-qualityengineeringculture

Silicon Valley's AI Civil War: Big Tech's Joint Letter vs. Anthropic's $40M Lobbying Push

Silicon Valley's AI Civil War: Big Tech's Joint Letter vs. Anthropic's $40M Lobbying Push

airegulationpolicyopen-source

AI in the Cockpit: An F-16 Just Took Its First Fully Autonomous Flight

AI in the Cockpit: An F-16 Just Took Its First Fully Autonomous Flight

AImilitaryDARPAethics

The $1.65 Trillion Debt Big Tech Doesn't Want You to See

The $1.65 Trillion Debt Big Tech Doesn't Want You to See

200 Startups to Trump: Don't Cut Us Off From Chinese Open-Source AI

200 Startups to Trump: Don't Cut Us Off From Chinese Open-Source AI

AIopen-sourceregulationChina

OpenAI's AI Model Broke Out of Its Sandbox, Hacked Hugging Face, and Stole Test Answers

OpenAI's AI Model Broke Out of Its Sandbox, Hacked Hugging Face, and Stole Test Answers

aisecurityopenai

Terence Tao, the World's Greatest Mathematician, Used ChatGPT to Crack the Jacobian Conjecture

Terence Tao, the World's Greatest Mathematician, Used ChatGPT to Crack the Jacobian Conjecture

aimathchatgpttao

$Outcounterexampled by Machines: How AI Toppled Three Major Conjectures in Two Months$

Outcounterexampled by Machines: How AI Toppled Three Major Conjectures in Two Months

aimathleanformal-verification

When AI Escaped the Sandbox: Inside the OpenAI–HuggingFace Security Incident

When AI Escaped the Sandbox: Inside the OpenAI–HuggingFace Security Incident

securityaiopenaihuggingface

AI-Generated Papers Surged From 2% to 15% on arXiv — and Academia Can't Keep Up

AI-Generated Papers Surged From 2% to 15% on arXiv — and Academia Can't Keep Up

AIAcademic IntegrityarXivResearch

AI Advice Made People 3x Less Accurate but 2x More Confident, Study Finds

AI Advice Made People 3x Less Accurate but 2x More Confident, Study Finds

AIResearchPsychologyCritical Thinking

China's AI Playbook: Giving Away the World's Smartest Models for Free

China's AI Playbook: Giving Away the World's Smartest Models for Free

AILLMOpen SourceUS-China Competition

OpenAI Silently Cuts Codex Context Window by 27% — Developers Cry 'Paid Downgrade'

OpenAI Silently Cuts Codex Context Window by 27% — Developers Cry 'Paid Downgrade'

AIOpenAIProgrammingCodex

$One Prompt, 148 Minutes: GPT-5.6 Closes a 30-Year Gap in Convex Optimization$

One Prompt, 148 Minutes: GPT-5.6 Closes a 30-Year Gap in Convex Optimization

AIGPT-5.6MathematicsFormal Proof

Six Months, No Tools: A Chinese AI Catches GPT-5.6

Six Months, No Tools: A Chinese AI Catches GPT-5.6

AIKimiChina AIGPT-5.6

The Collapse of Stack Overflow: Monthly Questions Down 99%

The Collapse of Stack Overflow: Monthly Questions Down 99%

Stack OverflowInternet HistoryAICommunity

Linux's Creator Pitched AI — and Got Rejected by His Own People

Linux's Creator Pitched AI — and Got Rejected by His Own People

LinuxLinus TorvaldsLLMOpen Source

A Chinese AI Beat GPT-5 in Benchmarks — and It's Fully Open Source

A Chinese AI Beat GPT-5 in Benchmarks — and It's Fully Open Source

Kimi K3MoonshotOpen SourceLarge Model

Linus Torvalds Pitched AI Code Review to the Kernel Team — and Got a Flat 'No'

Linus Torvalds Pitched AI Code Review to the Kernel Team — and Got a Flat 'No'

LinuxLinus TorvaldsLLMOpen Source

Not the Strongest, Yet Upvoted 559 Times: How American Open-Source AI Staged a Comeback

Not the Strongest, Yet Upvoted 559 Times: How American Open-Source AI Staged a Comeback

AIOpen SourceInklingLarge Model

A 13-Year-Old PC Runs Google's Latest AI — at 5 Tokens Per Second

A 13-Year-Old PC Runs Google's Latest AI — at 5 Tokens Per Second

AIGemmaCPU InferenceHardware

AI Is Rewriting How You Talk — and 405 People Lost It

AI Is Rewriting How You Talk — and 405 People Lost It

AILanguageClaudeWriting

353 Votes: Have You Outsourced Your Brain to AI Too?

353 Votes: Have You Outsourced Your Brain to AI Too?

AICognitive ScienceEducationThinking

Before It Even Reads Your Prompt, This AI Burns 33,000 Tokens

Before It Even Reads Your Prompt, This AI Burns 33,000 Tokens

AIClaude CodetokenBusiness Model

George Hotz: The Trillion-Dollar AI Valuation May Be a Mirage

George Hotz: The Trillion-Dollar AI Valuation May Be a Mirage

AIValuationBusiness ModelLLM

His Voice Was Cloned by AI. He's Had to Prove He's Human Five Times in One Year.

His Voice Was Cloned by AI. He's Had to Prove He's Human Five Times in One Year.

AIVoice SynthesisVoice ActingDeepfake

One Google Search, 100× Heavier: How the Web Got Fat on Carbon

One Google Search, 100× Heavier: How the Web Got Fat on Carbon

GoogleCarbon EmissionsEnvironmentInternet

Microsoft Fires idTech Engine Team: Why Xbox Is Dismantling Its Own Crown Jewel

Microsoft Fires idTech Engine Team: Why Xbox Is Dismantling Its Own Crown Jewel

MicrosoftGamingLayoffsid Software

Anthropic Finds a Hidden 'Broadcast Station' Inside Claude

Anthropic Finds a Hidden 'Broadcast Station' Inside Claude

AIInterpretabilityAnthropicClaude

GLM 5.2 Beats GPT-5.5 at 1/6 the Price — The AI Margin Collapse Has Begun

GLM 5.2 Beats GPT-5.5 at 1/6 the Price — The AI Margin Collapse Has Begun

AIGLMBusiness ModelOpen Source

How Claude and GPT Both Leaked Private Chats: The Shared Infrastructure Flaw Nobody's Fixing

How Claude and GPT Both Leaked Private Chats: The Shared Infrastructure Flaw Nobody's Fixing

AISecurityPrivacyClaude

How One Comment Can Steal Your Private YouTube Video Title: Google's AI Assistant Does the Rest

How One Comment Can Steal Your Private YouTube Video Title: Google's AI Assistant Does the Rest

YouTubePrivacySecurityAI

Claude Code Hid Steganographic Tracking in System Prompts — Here's How It Worked

Claude Code Hid Steganographic Tracking in System Prompts — Here's How It Worked

AIClaudeSecurityPrivacy

No Skull Drilling Required: Meta's Brain-to-Text AI Reaches 61% Accuracy

No Skull Drilling Required: Meta's Brain-to-Text AI Reaches 61% Accuracy

AIBrain-Computer InterfaceMetaNeuroscience

Anthropic Hid Invisible Watermarks in Every Claude Code Request — HN Exploded at 1,284 Points

Anthropic Hid Invisible Watermarks in Every Claude Code Request — HN Exploded at 1,284 Points

AISecurityPrivacySteganography

Alibaba's AI Now Runs on Your Laptop — No Internet Required

Alibaba's AI Now Runs on Your Laptop — No Internet Required

AIOpen SourceAlibabaLocal AI

Half the Class Used AI to Cheat. This Professor's Answer: Back to Paper and Pencil

Half the Class Used AI to Cheat. This Professor's Answer: Back to Paper and Pencil

AIEducationCheatingAcademic Integrity

$20 Beats Claude: How a Chinese Open-Source AI Pulled Off a Security Upset

$20 Beats Claude: How a Chinese Open-Source AI Pulled Off a Security Upset

AIOpen SourceGLMSecurity

AI Read My MRI and Said No Tear. The Doctor Glanced at It and Said Over 50% Torn

AI Read My MRI and Said No Tear. The Doctor Glanced at It and Said Over 50% Torn

AIHealthcareMRIDiagnosis

How AI Learned the "Dark Art" of RF Chip Design in One Week

How AI Learned the "Dark Art" of RF Chip Design in One Week

AIChip DesignRFICReinforcement Learning

Qualcomm's $3.9B Mojo Acquisition Opens a Software-Side Breach in CUDA's Moat

Qualcomm's $3.9B Mojo Acquisition Opens a Software-Side Breach in CUDA's Moat

AIAcquisitionCompiler TechnologySemiconductors

Krea 2 Open-Sourced: 12B Parameters Approaching Closed-Source SOTA

Krea 2 Open-Sourced: 12B Parameters Approaching Closed-Source SOTA

AIImage GenerationKreaOpen Source

The AI Free Lunch Is Almost Over

The AI Free Lunch Is Almost Over

AIeconomicsinference costsustainability

The Algorithmic Monoculture: When the Same AI Rejects You at Every Company

The Algorithmic Monoculture: When the Same AI Rejects You at Every Company

AIhiringalgorithmic biasmonoculture

Three Posts, One Day: The AI Coding Trust Crisis

Three Posts, One Day: The AI Coding Trust Crisis

AIcoding toolstrustCodex

Nature Study Confirms: Your AI Copilot Is Stealing Your Skills

Nature Study Confirms: Your AI Copilot Is Stealing Your Skills

AISkill DegradationCognitive ScienceHuman-AI Collaboration

Dango Daily · Daily tech trend briefing

Sources： Hacker News + Lobsters

Search