Notebooks Now in Gemini AppClaude Managed Agents Now in Public BetaCodex Retires Older Models on ChatGPT Login Apr 14Anthropic Launches Project Glasswing with Mythos Preview for Vuln DetectionAnthropic Secures Multi-GW TPU Deal with Google, BroadcomOpenAI Launches Safety Fellowship ProgramOpenAI Proposes Industrial Policy for Superintelligence EraCodex Biz/Ent Shift to Token-Based BillingComplete Guide to Building Skills for Claude ReleasedGoogle Cloud AI Agents Guide Explains Key DifferencesCodex Adds Vercel Plugin for Easy DeploymentClaude Subs No Longer Cover Third-Party ToolsClaude Subs Exclude 3rd-Party Tools, Go Pay-Per-UseAnthropic Releases Model Diffing for AI Behavior ComparisonNew Flex and Priority Tiers in Gemini APIGemma 4 Open Models ReleasedLyria 3 Prompting Tips in Gemini AppMicrosoft 365 Connectors Now on All Claude PlansGoogle Releases Gemma 4: Top Open Lightweight AI ModelClaude Computer Use expands to Windows (from macOS)Codex Pay-as-You-Go in ChatGPT Business, No Fixed Seat FeesGemini API Adds Flex & Priority TiersGoogle Vids Gains Veo 3.1 Video & Lyria Music Gen, Free TierEmotion vectors discovered in Claude causally affecting behaviorChatGPT Now on Apple CarPlay (iOS 26.4+)Google Releases Gemma 4 Open ModelsGoogle DeepMind Releases Gemma 4 Model Family (31B etc.)Linear Plugin Added to Codex AppSimplified Data Grounding in Gemini EnterpriseAnthropic Signs AI Safety MOU with AustraliaAnthropic Launches Claude Code Agentic Coding SystemClaude Code Leak Reveals Mythos ModelAnthropic Flattens High-Volume Pricing for Claude Opus/Sonnet APIPerplexity Scales Voice Search to Millions with Realtime APIOpenAI Codex Security Preview Remains FreeGKE Agent Sandbox Enables Sub-Second Agent StartsCodex Use Cases Gallery LaunchedGemini Personal Intelligence Free for US UsersCodex Plugins Enable Seamless Tool IntegrationDeepMind Releases AI Manipulation Measurement ToolkitOpenAI Reveals Model Spec Design and Evolution ProcessConversational Analytics Launches in BigQueryOpenAI Releases Teen Safety Prompts for gpt-oss-safeguardClaude Long-term Users Achieve 10% Higher Success Rate StudyMulti-Agent Harness Boosts Frontend DesignAnthropic Launches Science Blog for AI Science WorkflowsClaude Accelerates Grad-Level Physics CalcTrace Analytics Adds SQL Queries & ChartsCost-Effective LLM Serving with Ollama on GKE GPU SharingGoogle Adds MCP Servers for DBs, Boosting AI Agent IntegrationMigrate Apps to Cloud with Gemini CLI & MCP GuideDesign Delightful Frontends with GPT-5.4Parameter Golf Challenge for Minimal PromptsAnthropic Publishes 81k AI User Interviews (34 chars)Gemini 3 Preview: Thinking Levels & Image Gen Boost (42 chars)DeepMind Launches Cognitive Framework for AGI ProgressOpenAI Launches GPT-5.4 Mini & Nano Fast Small ModelsCloud Run Supports NVIDIA RTX PRO 6000 GPUs (38 chars)Sora API Adds Reusable Chars, 20s Videos, 1080pAnthropic Launches Institute for AI Societal ImpactsResponses API adds shell and container environmentChatGPT adds interactive math & science visualsOpenAI shares instruction-priority training approachOpenAI to acquire Promptfoo for eval securityGemini 3.1 Flash-Lite Preview: Fast, Low-Cost ModelUpdated GPT-5.4 Prompt Guide with Agent PatternsAnthropic Reveals Claude's Eval Awareness CaseAnthropic's Claude Opus 4.6 Finds 22 Firefox VulnerabilitiesAnthropic updates vulnerability disclosure policyAnthropic details a Claude exploit case studyGoogle ships TF 2.21, production LiteRT stackOpenAI Releases CoT Controllability Eval SuiteOpenAI Rolls Out GPT-5.4 Thinking & Pro in ChatGPTChatGPT for Excel beta and finance integrationsGPT-5.4 Thinking System Card details safety approachAnthropic proposes “observed exposure” labor metricGPT-5.4 pro arrives as Responses-only modelOpenAI ships a new GPT-5.4 API snapshotAnthropic explains current Department of War designationOpenAI boosts education support to close AI gapsOpenAI releases GPT‑5.4 for pro workflowsOpenAI releases new tools to measure learning outcomesCodex app adds Windows support rolloutGoogle ships February Gemini Drop updatesOpenAI publishes GPT‑5.3 Instant System CardGoogle adds Gemini in-app tasks in March Pixel DropClaude Skills Repo Open-SourcedOpenAI containers switch to session billing on 3/31OpenAI-DoW Safety Agreement for Classified AIAnthropic Challenges DoW Supply Risk LabelOpenAI-Amazon Partnership Launches Stateful Runtime on BedrockOpenAI Updates Assistants API Migration Guide with Sunset DateGoogle expands ADK integrations for agentsGoogle ships Nano Banana 2 in MENAOpenAI to add mental-health safety features soonGoogle expands ADK integrations for agentsAnthropic's Complete Guide to Claude SkillsOpenAI adds a stateful runtime in Amazon BedrockGoogle Releases Nano Banana 2 Prompt GuideCodex and Figma Integration Enables Auto UI Design from CodeOpenAI Announces End Date for Realtime API Beta: Feb 27, 2026OpenAI and Figma Launch Direct Codex-Design CollaborationAnthropic details defense talks in CEO statementOpenAI & PNNL aim to speed up permitting draftsAI Edge Gallery adds on-device function callingGoogle unveils Nano Banana 2 for faster Gemini imagesGoogle releases Nano Banana 2 for faster imagesGoogle Enhances Gemini 2.5 Audio Model for Live Voice AgentsAnthropic Acquires Vercept_ai for Claude Computer UseOpenAI Plans Realtime API Beta Shutdown on February 27, 2026Google updates Circle to Search for multi-object queriesOpenAI Publishes Report on AI Abuse Detection and MitigationGalaxy S26 Adds On-Device Gemini AI for Call Fraud DetectionAnthropic Publishes Responsible Scaling Policy v3.0 to Boost TransparencyOpenAI Launches GPT-5.3-Codex for Advanced Coding AgentsGoogle DeepMind Launches Robotics Accelerator for European StartupsAnthropic releases major RSP rewrite (v3.0)Anthropic releases RSP 3.0 governance rewriteAnthropic Updates Responsible Scaling Policy to Version 3.0Anthropic Publishes Responsible Scaling Policy 3.0 for Frontier AI SafetyGoogle Labs Adds Agent Procedures to Opal WorkflowsAnthropic Updates Responsible Scaling Policy to 3.0OpenAI Launches Frontier Alliances to Accelerate Enterprise AIAnthropic Publishes Claude AI Fluency Index MetricAnthropic details defenses vs distillation attacksOpenAI Plans to Retire SWE-bench Verified EvaluationAnthropic Updates Frontier Safety Roadmap with Feb 2026 GoalsAnthropic launches Claude Sonnet 4.6 with 1M contextChatGPT adds Lockdown Mode and risk labelsOpenAI launches Codex-Spark for ultra-low latencyGemini 3 Deep Think Gets Major Upgrade for ScienceOpenAI starts testing ads in ChatGPT (US)OpenAI launches GPT-5.3-Codex for agentsOpenAI Rolls Out Trusted Access for CybersecurityGoogle previews Developer Knowledge API + MCPXcode 26.3 Integrates Claude Agent SDK Nativelygpt-oss-120b Open Weights ReleasedCodex Integrates Work Context for PrioritiesCodex Hits 3M Weekly Users, Resets Rate Limits5 Insights for Agentic AI in 2026Claude Mythos Preview System Card ReleasedGoogle AI Studio Production UpdateGemma 4 runs offline on phones for local AI agentsGoogle Cloud Spotlights Gemini Embedding 2Claude Subs Exclude Third-Party ToolsClaude Subs Now Meter Third-Party Harness UsageClaude Subs Shift to Pay-Go for 3rd-Party ToolsGemini Provides Private Missed Summary in Google MeetVoice Agent Debugs Slides Live with gpt-realtime-1.5Codex Hooks Enable Lifecycle Script ExecutionGemini 3.1 Flash-Lite Released, Fastest & Cheapest with Audio BoostGoogle Vids Adds Free Veo 3.1 Video Gen & Lyria MusicOpenAI Acquires TBPN to Expand Global AI Media ConversationsBuild Coding Assistant with Gemini MCP/SkillsGemini 3.1 Flash-Lite Preview: 2.5x Faster at $0.25/M InputAnthropic Tightens Claude Subscription Limits During Peak HoursGemini API Adds Prepay Billing and Project Spend CapsGemini 3.1 Flash Live Preview Enhances Real-Time ChatStreet View Insights GA: Analyze 280B ImagesBigQuery Global Queries Across RegionsOpenAI Launches Safety Bug Bounty ProgramOpenAI Foundation Announces $1B Investment PlansGemini Automates Scheduling in Google CalendarPersonalized marketing campaigns with Vertex AIClaude Code advanced patterns webinar releasedGoogle Cloud shares 5 steps to scale AI beyond pilotsGKE LLM Inference Optimization QuickstartNeo4j Extension Added to Gemini CLI for NL Cypher QueriesGKE Inference Gateway Boosts Vertex AI LatencyOpenAI Acquires Astral to Boost CodexGemini API Adds Multi-Tool Calls & Google Maps GroundingSpanner CQL APIs Now GA on Google Cloud (32 chars)Gemini API Pricing Update Adds Gemini 3 Previews (38 chars)Gemini API Spend Caps Added to Google AI StudioBigQuery Studio Assistant Supercharged by GeminiBuild Onboarding Agents with Gemini Enterprise GuideClaude Now Creates Interactive Charts and DiagramsAnthropic Opens Sydney Office, Expands in APACOpenAI shares operational tips in Codex Prompting GuideGemini Sheets Hits SOTA on SpreadsheetBenchOpenAI Improves Instruction Hierarchy for SafetyGoogle adds Gemini scheduling help in GmailSkills accelerate OSS maintenance in Agents SDKCalendar updates now email as the principalOpenAI Academy updates ChatGPT Skills guideDescript: scaling multilingual dubbing with AIOpenAI shares an AI investing research case studyGoogle shares remixable I/O 2026 AI gamesClaude Opus 4.6 sabotage risk report publishedGPT-5.4 Thinking system card publishedChatGPT “Skills” guide updated for reusable workflowsGoogle Cloud Next '26 developer guide is outOpenAI Academy shares 4 GPT templates for SMBsOpenAI shares how Axios uses ChatGPT for journalismVeo 3 expands and reaches Gemini mobileOpenAI guide: structure prompts for caching winsOpenAI docs warn on safe MCP server connectionsGoogle explains how it built I/O 2026 with GeminiDeepMind shares Project Genie prompt tipsOpenAI documents Prompt Caching to cut latency and costOpenAI Loosens o1 Model Policy for Broader Commercial UseOpenAI Prompting guide covers versioned promptsAnthropic outlines prompt engineering basicsAnthropic shows Prompt improver workflowAnthropic Releases Code Gen Prompt Patterns for Claude 4Responses API migration guide for agent appsOpenAI explains prompt-based JSON schema generationGoogle AI Plus expands to 35 regions incl. USOpenAI Expands ChatGPT Enterprise to Japanese FirmsOpenAI guide for long-horizon Codex tasksAnthropic Details Prompt Improvement Steps for ClaudeGoogle Updates Basic Prompt Design Guide for GeminiOpenAI and Microsoft reaffirm partnership termsGoogle rolls out Nano Banana 2 across MENAGemini generates 30‑second songs with Lyria 3OpenAI and Microsoft reaffirm partnership termsGoogle publishes Gemini 3.1 Flash Image model cardGoogle Nano Banana 2 boosts fast image qualityGuide to Using System Instructions in Gemini APIOpenAI Updates Guide on Optimizing Prompt CachingAnthropic Updates Transparency Info for Claude Opus 4.6OpenAI Publishes PDF Summarizing Prompt Engineering BasicsOpenAI Details How to Use Prompt Caching EffectivelyOpenAI Academy ships prompt pack for gov tasksOpenAI Releases 'ChatGPT Tasks' Prompt Collection for GovernmentOpenAI Appoints Arvind KC as Chief People OfficerOpenAI Improves gpt-realtime-1.5 Voice Accuracy in Realtime APIGemini API Adds Multimodal Function Calls with Image SupportAnthropic explains Claude’s persona selection researchAnthropic Unveils Measures Against Distillation AttacksAnthropic previews Claude Code SecurityOpenAI publishes First Proof attempts with prompt appendixGoogle Releases Workflow for Creating XR 3D Experiences with GeminiGoogle Introduces Gemini's Lyria 3 Music Generation with Arabic BetaOpenAI Grants $7.5M to UK AISI’s Alignment ProjectOpenAI Releases EVMbench to Evaluate AI Agent Vulnerability DetectionGemini app adds Lyria 3 for music creationOpenAI Publishes California Privacy Requests DataOpenAI Warns About Unauthorized OpenAI Stock TradesConductor adds automated reviews for codeGemini CLI adds extension settings for easier setupTips for long‑running agents: Skills & ShellGemini CLI Extensions Simplify Config and Improve Key ManagementData Commons MCP becomes hosted for zero-install accessOpenAI brings ChatGPT to GenAI.milClaude Opus 4.6 boosts coding and long-context skillsAnthropic states Claude will remain ad-free in chatsGuide: Using Agentic Vision in Gemini 3 FlashGoogle Vertex AI Offers Prompt Templates for Scalable TestingGoogle upgrades Gemini and Classroom for educatorsGoogle Explains Key Prompt Design StrategiesAnthropic internal prompting leaks: constraint-focused styleBuild real-time agents with Gemini 3.1 Flash LiveGoogle AI Releases Vibe Coding Explainer VideoGemini in Gmail Enhances Data Privacy ExplainedClaude's Learning Mode Enables Step-by-Step TutoringCodex Workflow Building Spaces EventCustomize Gemini CLI with HooksClaude offers 13 free AI courses and certificatesOpen-Source Recreation of Claude Code Prompt ArchitectureCLAUDE.md Template for Claude Self-Improvement LoopsGemini Enterprise Automates Meeting PrepClaude Cowork Skill for Google Ads Dashboard CreationClaude Prompts Using MIT Prof Framework for PresentationsLearn Prompt Design & Tool Chaining from Leaked Claude Code SourceGoogle Offers Veo 3.1 Lite at Lowest Price in Gemini APIAnthropic announces AMA on Claude Code + MCPVertex AI Guide for Fine-Tuning Gemini 2.5 FlashUsing Codex to Migrate API to Latest ModelsData Grounding Guide Boosts Agent AccuracyGoogle Cloud Releases 101 GenAI BlueprintsNew Gemini/Claude Dashboards in Cloud Monitoring3 New Gemini Features Coming to Google TVSingle-tenant Cloud HSM for full key controlDataform adds hierarchical folder and repo structureSpanner MCP server codelab for natural language queriesOpenAI Updates Data Policy, Clarifies Training Opt-Out (44 chars)Claude 4.6 Makes 1M Token Context Standard PricingGenAI Podcasts from Audio Architecture GuideOpenAI Shares 1-Year Responses API Review & Use CasesWayfair improves catalog with OpenAI case studyGoogle launches Wednesday Build Hour for AI agent buildingAnthropic schedules agentic AI event (Mar 17)Google shares Lyria 3 music prompt guideGemini Image prompt guide for better resultsOpenAI shares a 5-step guide to AI adoptionGoogle posts the weekly Workspace updates recapPreprint: GPT-assisted quantum gravity resultAnthropic Updates Claude Connectors for Extended CapabilitiesAnthropic shares Claude 4 prompting best practicesGoogle refreshes prompt design intro for Vertex AIAnthropic Adds Image Upload to ClaudeGoogle Updates Data Analysis Prompts for GeminiAnthropic's AI Fluency Framework GuideElevenLabs expands Google Cloud tie-up for voice AIOpenAI Academy posts ChatGPT Tasks prompt pack for govOpenAI Academy publishes ChatGPT 102 workbookOpenAI Academy Shares Teacher Training Case StudyGoogle I/O 2026 set for May 19–20; Gemini teasedGoogle showcases practical Gemini use cases in new adsAnthropic publishes a PDF of Claude’s ConstitutionGoogle releases TranslateGemma open translation modelsOpenAI Releases GPT-5.1 Prompting Guide w/ New ToolsNotebooks Now in Gemini AppClaude Managed Agents Now in Public BetaCodex Retires Older Models on ChatGPT Login Apr 14Anthropic Launches Project Glasswing with Mythos Preview for Vuln DetectionAnthropic Secures Multi-GW TPU Deal with Google, BroadcomOpenAI Launches Safety Fellowship ProgramOpenAI Proposes Industrial Policy for Superintelligence EraCodex Biz/Ent Shift to Token-Based BillingComplete Guide to Building Skills for Claude ReleasedGoogle Cloud AI Agents Guide Explains Key DifferencesCodex Adds Vercel Plugin for Easy DeploymentClaude Subs No Longer Cover Third-Party ToolsClaude Subs Exclude 3rd-Party Tools, Go Pay-Per-UseAnthropic Releases Model Diffing for AI Behavior ComparisonNew Flex and Priority Tiers in Gemini APIGemma 4 Open Models ReleasedLyria 3 Prompting Tips in Gemini AppMicrosoft 365 Connectors Now on All Claude PlansGoogle Releases Gemma 4: Top Open Lightweight AI ModelClaude Computer Use expands to Windows (from macOS)Codex Pay-as-You-Go in ChatGPT Business, No Fixed Seat FeesGemini API Adds Flex & Priority TiersGoogle Vids Gains Veo 3.1 Video & Lyria Music Gen, Free TierEmotion vectors discovered in Claude causally affecting behaviorChatGPT Now on Apple CarPlay (iOS 26.4+)Google Releases Gemma 4 Open ModelsGoogle DeepMind Releases Gemma 4 Model Family (31B etc.)Linear Plugin Added to Codex AppSimplified Data Grounding in Gemini EnterpriseAnthropic Signs AI Safety MOU with AustraliaAnthropic Launches Claude Code Agentic Coding SystemClaude Code Leak Reveals Mythos ModelAnthropic Flattens High-Volume Pricing for Claude Opus/Sonnet APIPerplexity Scales Voice Search to Millions with Realtime APIOpenAI Codex Security Preview Remains FreeGKE Agent Sandbox Enables Sub-Second Agent StartsCodex Use Cases Gallery LaunchedGemini Personal Intelligence Free for US UsersCodex Plugins Enable Seamless Tool IntegrationDeepMind Releases AI Manipulation Measurement ToolkitOpenAI Reveals Model Spec Design and Evolution ProcessConversational Analytics Launches in BigQueryOpenAI Releases Teen Safety Prompts for gpt-oss-safeguardClaude Long-term Users Achieve 10% Higher Success Rate StudyMulti-Agent Harness Boosts Frontend DesignAnthropic Launches Science Blog for AI Science WorkflowsClaude Accelerates Grad-Level Physics CalcTrace Analytics Adds SQL Queries & ChartsCost-Effective LLM Serving with Ollama on GKE GPU SharingGoogle Adds MCP Servers for DBs, Boosting AI Agent IntegrationMigrate Apps to Cloud with Gemini CLI & MCP GuideDesign Delightful Frontends with GPT-5.4Parameter Golf Challenge for Minimal PromptsAnthropic Publishes 81k AI User Interviews (34 chars)Gemini 3 Preview: Thinking Levels & Image Gen Boost (42 chars)DeepMind Launches Cognitive Framework for AGI ProgressOpenAI Launches GPT-5.4 Mini & Nano Fast Small ModelsCloud Run Supports NVIDIA RTX PRO 6000 GPUs (38 chars)Sora API Adds Reusable Chars, 20s Videos, 1080pAnthropic Launches Institute for AI Societal ImpactsResponses API adds shell and container environmentChatGPT adds interactive math & science visualsOpenAI shares instruction-priority training approachOpenAI to acquire Promptfoo for eval securityGemini 3.1 Flash-Lite Preview: Fast, Low-Cost ModelUpdated GPT-5.4 Prompt Guide with Agent PatternsAnthropic Reveals Claude's Eval Awareness CaseAnthropic's Claude Opus 4.6 Finds 22 Firefox VulnerabilitiesAnthropic updates vulnerability disclosure policyAnthropic details a Claude exploit case studyGoogle ships TF 2.21, production LiteRT stackOpenAI Releases CoT Controllability Eval SuiteOpenAI Rolls Out GPT-5.4 Thinking & Pro in ChatGPTChatGPT for Excel beta and finance integrationsGPT-5.4 Thinking System Card details safety approachAnthropic proposes “observed exposure” labor metricGPT-5.4 pro arrives as Responses-only modelOpenAI ships a new GPT-5.4 API snapshotAnthropic explains current Department of War designationOpenAI boosts education support to close AI gapsOpenAI releases GPT‑5.4 for pro workflowsOpenAI releases new tools to measure learning outcomesCodex app adds Windows support rolloutGoogle ships February Gemini Drop updatesOpenAI publishes GPT‑5.3 Instant System CardGoogle adds Gemini in-app tasks in March Pixel DropClaude Skills Repo Open-SourcedOpenAI containers switch to session billing on 3/31OpenAI-DoW Safety Agreement for Classified AIAnthropic Challenges DoW Supply Risk LabelOpenAI-Amazon Partnership Launches Stateful Runtime on BedrockOpenAI Updates Assistants API Migration Guide with Sunset DateGoogle expands ADK integrations for agentsGoogle ships Nano Banana 2 in MENAOpenAI to add mental-health safety features soonGoogle expands ADK integrations for agentsAnthropic's Complete Guide to Claude SkillsOpenAI adds a stateful runtime in Amazon BedrockGoogle Releases Nano Banana 2 Prompt GuideCodex and Figma Integration Enables Auto UI Design from CodeOpenAI Announces End Date for Realtime API Beta: Feb 27, 2026OpenAI and Figma Launch Direct Codex-Design CollaborationAnthropic details defense talks in CEO statementOpenAI & PNNL aim to speed up permitting draftsAI Edge Gallery adds on-device function callingGoogle unveils Nano Banana 2 for faster Gemini imagesGoogle releases Nano Banana 2 for faster imagesGoogle Enhances Gemini 2.5 Audio Model for Live Voice AgentsAnthropic Acquires Vercept_ai for Claude Computer UseOpenAI Plans Realtime API Beta Shutdown on February 27, 2026Google updates Circle to Search for multi-object queriesOpenAI Publishes Report on AI Abuse Detection and MitigationGalaxy S26 Adds On-Device Gemini AI for Call Fraud DetectionAnthropic Publishes Responsible Scaling Policy v3.0 to Boost TransparencyOpenAI Launches GPT-5.3-Codex for Advanced Coding AgentsGoogle DeepMind Launches Robotics Accelerator for European StartupsAnthropic releases major RSP rewrite (v3.0)Anthropic releases RSP 3.0 governance rewriteAnthropic Updates Responsible Scaling Policy to Version 3.0Anthropic Publishes Responsible Scaling Policy 3.0 for Frontier AI SafetyGoogle Labs Adds Agent Procedures to Opal WorkflowsAnthropic Updates Responsible Scaling Policy to 3.0OpenAI Launches Frontier Alliances to Accelerate Enterprise AIAnthropic Publishes Claude AI Fluency Index MetricAnthropic details defenses vs distillation attacksOpenAI Plans to Retire SWE-bench Verified EvaluationAnthropic Updates Frontier Safety Roadmap with Feb 2026 GoalsAnthropic launches Claude Sonnet 4.6 with 1M contextChatGPT adds Lockdown Mode and risk labelsOpenAI launches Codex-Spark for ultra-low latencyGemini 3 Deep Think Gets Major Upgrade for ScienceOpenAI starts testing ads in ChatGPT (US)OpenAI launches GPT-5.3-Codex for agentsOpenAI Rolls Out Trusted Access for CybersecurityGoogle previews Developer Knowledge API + MCPXcode 26.3 Integrates Claude Agent SDK Nativelygpt-oss-120b Open Weights ReleasedCodex Integrates Work Context for PrioritiesCodex Hits 3M Weekly Users, Resets Rate Limits5 Insights for Agentic AI in 2026Claude Mythos Preview System Card ReleasedGoogle AI Studio Production UpdateGemma 4 runs offline on phones for local AI agentsGoogle Cloud Spotlights Gemini Embedding 2Claude Subs Exclude Third-Party ToolsClaude Subs Now Meter Third-Party Harness UsageClaude Subs Shift to Pay-Go for 3rd-Party ToolsGemini Provides Private Missed Summary in Google MeetVoice Agent Debugs Slides Live with gpt-realtime-1.5Codex Hooks Enable Lifecycle Script ExecutionGemini 3.1 Flash-Lite Released, Fastest & Cheapest with Audio BoostGoogle Vids Adds Free Veo 3.1 Video Gen & Lyria MusicOpenAI Acquires TBPN to Expand Global AI Media ConversationsBuild Coding Assistant with Gemini MCP/SkillsGemini 3.1 Flash-Lite Preview: 2.5x Faster at $0.25/M InputAnthropic Tightens Claude Subscription Limits During Peak HoursGemini API Adds Prepay Billing and Project Spend CapsGemini 3.1 Flash Live Preview Enhances Real-Time ChatStreet View Insights GA: Analyze 280B ImagesBigQuery Global Queries Across RegionsOpenAI Launches Safety Bug Bounty ProgramOpenAI Foundation Announces $1B Investment PlansGemini Automates Scheduling in Google CalendarPersonalized marketing campaigns with Vertex AIClaude Code advanced patterns webinar releasedGoogle Cloud shares 5 steps to scale AI beyond pilotsGKE LLM Inference Optimization QuickstartNeo4j Extension Added to Gemini CLI for NL Cypher QueriesGKE Inference Gateway Boosts Vertex AI LatencyOpenAI Acquires Astral to Boost CodexGemini API Adds Multi-Tool Calls & Google Maps GroundingSpanner CQL APIs Now GA on Google Cloud (32 chars)Gemini API Pricing Update Adds Gemini 3 Previews (38 chars)Gemini API Spend Caps Added to Google AI StudioBigQuery Studio Assistant Supercharged by GeminiBuild Onboarding Agents with Gemini Enterprise GuideClaude Now Creates Interactive Charts and DiagramsAnthropic Opens Sydney Office, Expands in APACOpenAI shares operational tips in Codex Prompting GuideGemini Sheets Hits SOTA on SpreadsheetBenchOpenAI Improves Instruction Hierarchy for SafetyGoogle adds Gemini scheduling help in GmailSkills accelerate OSS maintenance in Agents SDKCalendar updates now email as the principalOpenAI Academy updates ChatGPT Skills guideDescript: scaling multilingual dubbing with AIOpenAI shares an AI investing research case studyGoogle shares remixable I/O 2026 AI gamesClaude Opus 4.6 sabotage risk report publishedGPT-5.4 Thinking system card publishedChatGPT “Skills” guide updated for reusable workflowsGoogle Cloud Next '26 developer guide is outOpenAI Academy shares 4 GPT templates for SMBsOpenAI shares how Axios uses ChatGPT for journalismVeo 3 expands and reaches Gemini mobileOpenAI guide: structure prompts for caching winsOpenAI docs warn on safe MCP server connectionsGoogle explains how it built I/O 2026 with GeminiDeepMind shares Project Genie prompt tipsOpenAI documents Prompt Caching to cut latency and costOpenAI Loosens o1 Model Policy for Broader Commercial UseOpenAI Prompting guide covers versioned promptsAnthropic outlines prompt engineering basicsAnthropic shows Prompt improver workflowAnthropic Releases Code Gen Prompt Patterns for Claude 4Responses API migration guide for agent appsOpenAI explains prompt-based JSON schema generationGoogle AI Plus expands to 35 regions incl. USOpenAI Expands ChatGPT Enterprise to Japanese FirmsOpenAI guide for long-horizon Codex tasksAnthropic Details Prompt Improvement Steps for ClaudeGoogle Updates Basic Prompt Design Guide for GeminiOpenAI and Microsoft reaffirm partnership termsGoogle rolls out Nano Banana 2 across MENAGemini generates 30‑second songs with Lyria 3OpenAI and Microsoft reaffirm partnership termsGoogle publishes Gemini 3.1 Flash Image model cardGoogle Nano Banana 2 boosts fast image qualityGuide to Using System Instructions in Gemini APIOpenAI Updates Guide on Optimizing Prompt CachingAnthropic Updates Transparency Info for Claude Opus 4.6OpenAI Publishes PDF Summarizing Prompt Engineering BasicsOpenAI Details How to Use Prompt Caching EffectivelyOpenAI Academy ships prompt pack for gov tasksOpenAI Releases 'ChatGPT Tasks' Prompt Collection for GovernmentOpenAI Appoints Arvind KC as Chief People OfficerOpenAI Improves gpt-realtime-1.5 Voice Accuracy in Realtime APIGemini API Adds Multimodal Function Calls with Image SupportAnthropic explains Claude’s persona selection researchAnthropic Unveils Measures Against Distillation AttacksAnthropic previews Claude Code SecurityOpenAI publishes First Proof attempts with prompt appendixGoogle Releases Workflow for Creating XR 3D Experiences with GeminiGoogle Introduces Gemini's Lyria 3 Music Generation with Arabic BetaOpenAI Grants $7.5M to UK AISI’s Alignment ProjectOpenAI Releases EVMbench to Evaluate AI Agent Vulnerability DetectionGemini app adds Lyria 3 for music creationOpenAI Publishes California Privacy Requests DataOpenAI Warns About Unauthorized OpenAI Stock TradesConductor adds automated reviews for codeGemini CLI adds extension settings for easier setupTips for long‑running agents: Skills & ShellGemini CLI Extensions Simplify Config and Improve Key ManagementData Commons MCP becomes hosted for zero-install accessOpenAI brings ChatGPT to GenAI.milClaude Opus 4.6 boosts coding and long-context skillsAnthropic states Claude will remain ad-free in chatsGuide: Using Agentic Vision in Gemini 3 FlashGoogle Vertex AI Offers Prompt Templates for Scalable TestingGoogle upgrades Gemini and Classroom for educatorsGoogle Explains Key Prompt Design StrategiesAnthropic internal prompting leaks: constraint-focused styleBuild real-time agents with Gemini 3.1 Flash LiveGoogle AI Releases Vibe Coding Explainer VideoGemini in Gmail Enhances Data Privacy ExplainedClaude's Learning Mode Enables Step-by-Step TutoringCodex Workflow Building Spaces EventCustomize Gemini CLI with HooksClaude offers 13 free AI courses and certificatesOpen-Source Recreation of Claude Code Prompt ArchitectureCLAUDE.md Template for Claude Self-Improvement LoopsGemini Enterprise Automates Meeting PrepClaude Cowork Skill for Google Ads Dashboard CreationClaude Prompts Using MIT Prof Framework for PresentationsLearn Prompt Design & Tool Chaining from Leaked Claude Code SourceGoogle Offers Veo 3.1 Lite at Lowest Price in Gemini APIAnthropic announces AMA on Claude Code + MCPVertex AI Guide for Fine-Tuning Gemini 2.5 FlashUsing Codex to Migrate API to Latest ModelsData Grounding Guide Boosts Agent AccuracyGoogle Cloud Releases 101 GenAI BlueprintsNew Gemini/Claude Dashboards in Cloud Monitoring3 New Gemini Features Coming to Google TVSingle-tenant Cloud HSM for full key controlDataform adds hierarchical folder and repo structureSpanner MCP server codelab for natural language queriesOpenAI Updates Data Policy, Clarifies Training Opt-Out (44 chars)Claude 4.6 Makes 1M Token Context Standard PricingGenAI Podcasts from Audio Architecture GuideOpenAI Shares 1-Year Responses API Review & Use CasesWayfair improves catalog with OpenAI case studyGoogle launches Wednesday Build Hour for AI agent buildingAnthropic schedules agentic AI event (Mar 17)Google shares Lyria 3 music prompt guideGemini Image prompt guide for better resultsOpenAI shares a 5-step guide to AI adoptionGoogle posts the weekly Workspace updates recapPreprint: GPT-assisted quantum gravity resultAnthropic Updates Claude Connectors for Extended CapabilitiesAnthropic shares Claude 4 prompting best practicesGoogle refreshes prompt design intro for Vertex AIAnthropic Adds Image Upload to ClaudeGoogle Updates Data Analysis Prompts for GeminiAnthropic's AI Fluency Framework GuideElevenLabs expands Google Cloud tie-up for voice AIOpenAI Academy posts ChatGPT Tasks prompt pack for govOpenAI Academy publishes ChatGPT 102 workbookOpenAI Academy Shares Teacher Training Case StudyGoogle I/O 2026 set for May 19–20; Gemini teasedGoogle showcases practical Gemini use cases in new adsAnthropic publishes a PDF of Claude’s ConstitutionGoogle releases TranslateGemma open translation modelsOpenAI Releases GPT-5.1 Prompting Guide w/ New Tools
🔒 公式発表のみ掲載。噂・リーク・情報商材は載せません。
← Back to glossary

Zero-shot Prompting

ゼロショットプロンプティング

ぜろしょっとぷろんぷてぃんぐ

Definition

Zero-shot prompting is a method that asks a model to perform a task using only instructions, without providing examples. It is useful for quick trials and simple tasks.

LLMに何かを頼むとき、多くの人は無意識にゼロショットプロンプティングを実践しています。ゼロショットプロンプティングとは、具体的な入出力例を一切示さずに、指示文だけでモデルにタスクを実行させるプロンプト手法です。

最もシンプルなプロンプト手法

「この文章を要約してください」「以下のレビューがポジティブかネガティブか判定してください」——これらはすべてゼロショットプロンプティングです。求めるタスクの説明だけを与え、「こういう入力にはこう返してほしい」という例示を一切含めません。LLMが事前学習で獲得した汎用的な知識と推論能力に全面的に頼るアプローチです。

ゼロショットが有効な場面

ゼロショットプロンプティングは、タスクが一般的で、期待する出力形式が自明な場合に最も効果を発揮します。翻訳、要約、感情分析、質問応答といった広く知られたNLPタスクでは、モデルが学習時に大量の類似タスクに触れているため、例を示さなくても高い精度が期待できます。新しいタスクをすばやく試す際のプロトタイピングにも最適です。プロンプトが短いため入力トークン数が少なく、コスト面でも有利です。

限界を知る

しかし、ゼロショットでは対応しきれない場面も多くあります。独自のフォーマットで出力してほしい場合(たとえば社内独自のレポート形式)、判断基準が暗黙的な場合(「良い」コードレビューコメントとは何か)、あるいはモデルが不慣れなドメイン固有のタスクでは、精度が大きく低下します。出力の一貫性が重要な本番環境では、同じプロンプトでも回答のばらつきが大きくなりやすい点も課題です。

Few-shotへの移行判断

ゼロショットで試してみて結果が不十分な場合、次に検討すべきはFew-shotプロンプティングです。判断基準はシンプルで、出力のフォーマットが安定しない、判断基準を言葉だけで説明しきれない、エッジケースでの挙動を制御したい——これらに該当するなら、2〜3個の具体例を加えることで劇的に改善する場合があります。逆に、ゼロショットで十分な結果が出ているなら、わざわざ例を追加してトークンを消費する必要はありません。

効果を高めるコツ

ゼロショットでも、指示の書き方次第で精度は大きく変わります。「要約して」ではなく「3文以内で、技術者向けに、主要な変更点に焦点を当てて要約してください」のように、対象読者・長さ・焦点を明示することがポイントです。また、出力の形式(JSON、箇条書き、表など)を指定するだけでも、回答の品質と一貫性が向上します。

h
hayami

Stay on top of OpenAI, Google & Anthropic updates. An AI digest for business professionals.

Source Policy

We use only official sources. Each article links to the original announcement so you can verify it yourself.

© 2026 hayami. All rights reserved.