Dateline: May 22, 2026 | Next update: May 29, 2026
Another dense week anchored by three headline stories: Andrej Karpathy (OpenAI co-founder, Tesla AI director) joined Anthropic's pre-training team on May 19; self-hosted sandboxes and MCP tunnels launched for Claude Managed Agents at the Code with Claude London event on the same day; and the Gates Foundation — already announced May 14 — became fully detailed as Anthropic's largest mission-aligned partnership ($200M over four years). PwC expanded its Anthropic alliance to certify 30,000 professionals on Claude, and the API gained cache diagnostics in public beta.
Claude / Anthropic
Andrej Karpathy joins Anthropic — pre-training team
Announced: May 19, 2026 | Role: Member of Technical Staff, pre-training | Reports to: Nick Joseph
Andrej Karpathy — OpenAI founding team member, former Tesla senior director of AI, and the most widely followed AI educator in the world — joined Anthropic on May 19. He will work under pre-training team lead Nick Joseph with a specific mandate: start a new team that uses Claude itself to accelerate pre-training research. This is a direct bet that AI-assisted experimentation — running more ablations, summarising internal results, writing infrastructure code, and proposing training recipes — is the next competitive lever in the frontier model race.
★ What's newKarpathy starts at Anthropic on May 19. He will lead a new group within pre-training focused on using Claude to accelerate the training research loop. He paused his education startup Eureka Labs to join. He had been publicly enthusiastic about Claude Code since January 2026. His hire comes as Anthropic's annualised revenue run rate reaches $30B and Claude Code alone generates approximately $2.5B in annualised revenue.
Technical detailsTeam: pre-training | Lead: Nick Joseph | Mission: AI-accelerated pre-training research | Background: OpenAI founding (2015), Tesla Autopilot (2017–2022), OpenAI (2023), Eureka Labs (2024–2026) | Education work: plans to resume in time
Best for: Watch this space — Karpathy's focus on AI-accelerated pre-training research is a long-term signal, not an immediate product change
Claude Managed Agents — self-hosted sandboxes + MCP tunnels
Announced: May 19, 2026 at Code with Claude London | Self-hosted sandboxes: public beta | MCP tunnels: research preview (request access)
Announced at Anthropic's first international developer conference — Code with Claude London — self-hosted sandboxes and MCP tunnels together solve the core blocker for enterprise agent production deployments: data and code leaving the customer's security perimeter. The agent orchestration loop stays on Anthropic's infrastructure; tool execution and private service access move inside the customer's perimeter.
★ What's newSelf-hosted sandboxes (public beta): tool execution now runs on your own infrastructure or through managed providers — Cloudflare, Daytona, Modal, or Vercel. Your network policies, audit logging, and security tooling apply by default. Resource sizing and runtime images are fully customer-controlled. MCP tunnels (research preview): a lightweight gateway you deploy makes a single outbound encrypted connection — no inbound firewall rules, no public endpoints. Agents reach internal databases, private APIs, ticketing systems, and knowledge bases as MCP tools without those services being exposed to the public internet. Also shipping: MCP server and tool configurations can now be updated on an active session (no restart required); large tool outputs over 100K tokens are automatically spilled to a sandbox file, with the model receiving a truncated preview and the path to read the full content.
Technical detailsSelf-hosted sandboxes: public beta | Supported providers: Cloudflare (microVMs), Daytona, Modal, Vercel | MCP tunnels: research preview, request access via Claude Console workspace settings | Orchestration stays on Anthropic infra | Live MCP config updates: yes, no session restart required | Large output spill: >100K tokens auto-spilled to sandbox file | Both under managed-agents-2026-04-01 header
Best for: Enterprises in regulated industries (finance, healthcare, legal, government) requiring agent execution inside their own security perimeter
Gates Foundation — $200M four-year partnership
Announced: May 14, 2026 | Commitment: $200M over four years | Focus: global health, education, economic mobility
Anthropic committed $200 million over four years — in grant funding, Claude usage credits, and embedded engineering support — to a partnership with the Bill and Melinda Gates Foundation. It is the Gates Foundation's largest AI commitment to date (4x the $50M it gave OpenAI in January for a narrower African clinic programme). The work is led by Anthropic's Beneficial Deployments team.
★ What's newHealthcare focus: accelerating vaccine and drug candidate screening for neglected diseases (including polio, HPV, preeclampsia), and improving malaria and tuberculosis treatment deployment forecasts with the Institute for Disease Modeling. The partnership will serve the approximately 4.6 billion people in low- and middle-income countries who lack access to essential health services. Education focus: tools to improve K–12 outcomes in the US, sub-Saharan Africa, and India, including African-language data collection and labelling released as public goods. Economic mobility: support for solopreneur and small-business programmes. Public goods deliverables include open health datasets, evaluation benchmarks, and knowledge graphs.
Technical detailsCommitment: $200M over 4 years | Components: grant funding + Claude usage credits + Anthropic engineering support | Led by: Anthropic Beneficial Deployments team | Partner: Bill and Melinda Gates Foundation | Public goods: open datasets, benchmarks, knowledge graphs released to the broader research community | Comparison: Gates Foundation gave OpenAI $50M in Jan 2026 for narrower African clinic work
Best for: Global health researchers, educators, nonprofits, and policymakers — informational for commercial users
PwC — expanded strategic alliance
Announced: May 14, 2026 | Scope: global PwC workforce
Anthropic and PwC disclosed a major expansion of their strategic alliance. The rollout starts with PwC's US teams and extends to hundreds of thousands of professionals globally, anchored by a joint Center of Excellence and a certification programme targeting 30,000 PwC professionals trained and certified on Claude.
★ What's newClaude Code and Claude Cowork rolling out across PwC's global workforce. Joint Anthropic–PwC Center of Excellence established. 30,000 PwC professionals to be trained and certified on Claude. Deployment begins with US teams, expanding globally. This makes PwC one of the largest single enterprise deployments of Claude to date.
Technical detailsProducts: Claude Code + Claude Cowork | Starting geography: US, expanding globally | Center of Excellence: joint Anthropic–PwC | Certification target: 30,000 professionals
Best for: Enterprise buyers evaluating large-scale professional services deployment; consulting and advisory firms
API — cache diagnostics in public beta
Platform: Claude API | Beta header: cache-diagnosis-2026-04-07 | Availability: all API users
Prompt caching is one of the most impactful cost-reduction tools on the Claude API, but until this week there was no way to know why a cache miss happened. Cache diagnostics, now in public beta, tells you exactly where the prompt cache prefix diverged from the previous turn — making it possible to debug and fix caching issues without guesswork.
★ What's newPass diagnostics.previous_message_id on a Messages request and the API returns a cache_miss_reason field explaining where the cache prefix diverged. Enables systematic debugging of prompt cache misses across multi-turn agentic workflows.
Technical detailsBeta header: cache-diagnosis-2026-04-07 | Parameter: diagnostics.previous_message_id | Response field: cache_miss_reason | Works on: Messages API | Use case: debug prompt cache misses in agentic and multi-turn workflows
Best for: API developers optimising cost and latency through prompt caching
Claude Code — Agent Teams, plugin fixes, session tools
Platform: terminal / VS Code / web / mobile | Availability: all plans
Claude Code shipped a broad point release focused on Agent Teams stability, plugin improvements, and session tooling polish. Agent Teams — collaborative multi-agent sessions where named teammates work together on a task — received several reliability fixes after non-ASCII names caused API failures and crashed sessions.
★ What's newAgent Teams fixes: teammates with non-ASCII names no longer fail every API call due to invalid header encoding. Plugin marketplace: add/update now respects CLAUDE_CODE_PLUGIN_PREFER_HTTPS; /plugin returns to the Installed list after enable/disable/uninstall actions. /doctor now shows an exec-form example when a command hook is missing the command field. Skill-listing truncation moved out of startup notifications — run /doctor for the full breakdown. Pre-response stream stall recovery improved — now retries streaming once rather than falling back to the slower non-streaming path. SDK/headless MCP startup is up to 2 seconds faster with slow MCP servers (pre-wait now overlaps startup instead of blocking). Fixed: infinite loop where a skill using context: fork repeatedly re-invoked itself. Fixed: /review using a deprecated projectCards GraphQL query that errored on repos with Classic Projects.
Technical detailsAgent Teams: non-ASCII name header encoding fixed | CLAUDE_CODE_PLUGIN_PREFER_HTTPS: now respected on marketplace add/update | MCP headless startup: up to 2s faster | Stream stall: retry-streaming-once before fallback | Fixed: context: fork infinite loop | Fixed: /review GraphQL projectCards deprecation | Fixed: stale 'Failed to install Anthropic marketplace' banner | Fixed: PR badge not updating after gh pr create in-session
Best for: Developers using Agent Teams, MCP servers in headless SDK mode, plugin marketplace
Usage limits — Pro and Max tightening reported
Reported: May 14, 2026 | Affects: Pro and Max consumer plans
Anthropic quietly tightened Claude usage limits for Pro and Max subscribers. The API was not affected. This is a consumer-plan change only and came shortly after the doubling of Claude Code five-hour limits via the SpaceX compute deal announced May 6. Anthropic has not published a detailed breakdown of the new limits.
★ What's newPro and Max plan usage limits tightened as of approximately May 14. API pricing and limits unchanged. The net effect on Claude Code five-hour limits relative to the May 6 doubling is unclear — monitor your usage dashboard if you are a heavy Pro or Max user.
Technical detailsAffected: Pro ($20/mo) and Max ($100–200/mo) consumer plans | API: unaffected | Source: Axios, May 14, 2026 | No official Anthropic statement on specifics | Practical impact: most visible on long Claude Code sessions without context auto-compaction
Best for: Heavy Pro and Max subscribers — check your usage dashboard; no action needed for API users
Plans and Pricing
No API pricing changes this week. The usage limit tightening on Pro and Max plans is the only consumer-facing change and has not been quantified officially. Claude Platform on AWS consumption pricing via AWS Marketplace remains the same as native Claude API rates.
Technical detailsOpus 4.7: $5/$25 per MTok | Sonnet 4.6: $3/$15 per MTok | Haiku 4.5: low-cost tier | Claude Platform on AWS: consumption via AWS Marketplace | Gates Foundation and PwC: enterprise/mission-aligned pricing, contact Anthropic | Self-hosted sandboxes + MCP tunnels: included in Managed Agents (beta)
ChatGPT / OpenAI
Dateline: May 22, 2026 | Next update: May 29, 2026
Over the past week, OpenAI has continued improving agent reliability, multimodal workflows, and enterprise integration, while quietly refining system consistency across professional use cases.
GPT-5.3 Standard — default model
Release: late 2025 | Pricing: included | Availability: all users
More reliable handling of long and mixed-format conversations.
★ What's newImproved consistency when switching between text, document, and image-based tasks.
Technical detailsContext ~128k | Output ~4k–8k | Improved multimodal context management
Best for: General use
GPT-5.3 Pro — high-reasoning model
Release: late 2025 | Pricing: Pro | Availability: Pro/Enterprise
More stable performance on layered analytical tasks.
★ What's newImproved reasoning continuity in very long outputs.
Technical detailsContext ~200k (est.) | Reduced degradation across extended reasoning chains
Best for: Deep analysis
GPT-5.3 Mini — fallback model
Release: late 2025 | Pricing: low-cost | Availability: all
Faster and smoother lightweight responses.
★ What's newImproved routing precision between Mini and Standard models.
Technical detailsContext ~64k | Better dynamic inference allocation
Best for: Quick tasks
Agent Mode
Handles multi-step workflows.
★ What's newImproved reliability in executing structured workflows over longer sessions.
Technical detailsBetter task persistence | Reduced workflow interruption rates
Best for: Task delegation
Deep Research
Combines browsing and reasoning.
★ What's newImproved handling of conflicting information across multiple sources.
Technical detailsEnhanced synthesis and source-weighting pipeline
Best for: Research
Memory & Projects
Persistent context across chats.
★ What's newBetter prioritisation of active-project context over older conversational memory.
Technical detailsImproved relevance filtering | More efficient memory retrieval
Best for: Ongoing workflows
Advanced Voice Mode
★ What's newMore natural transitions between conversational turns.
Technical detailsImproved turn-taking latency | Better conversational pacing
Best for: Voice interaction
ChatGPT for Clinicians
Continued expansion of healthcare-oriented workflows.
★ What's newImproved structure and readability in generated clinical summaries and administrative outputs.
Technical detailsFurther domain-specific tuning | Reinforced medical safety guardrails
Best for: Clinical support (non-diagnostic assistance)
Enterprise & Workflow Integrations
★ What's newImproved reliability across integrations involving documents, collaborative workflows, and persistent projects. Better orchestration across connected tools and memory systems.
Best for: Organisational deployment
Plans and Pricing
No significant changes this week. Pricing stable | API structure unchanged.
Gemini (Google)
Date: May 22, 2026 | Next update: May 29, 2026
This week marked the kickoff of Google I/O 2026, officially shifting the ecosystem into the "Agentic Gemini Era." Google introduced a brand-new model generation, multimodal video foundational architectures, and autonomous cloud agents.
Gemini 3.5 Flash — new default model
Launched: Google I/O 2026 | Availability: Gemini app and Google Search (default)
Launched as the default engine across the Gemini app and Google Search. Runs four times faster than other frontier models in output tokens per second and outperforms Gemini 3.1 Pro across key logic and coding benchmarks.
★ What's newGemini 3.5 Flash is now the default model across the Gemini app and Google Search. 4x faster output token throughput than competing frontier models. Outperforms Gemini 3.1 Pro on logic and coding benchmarks. Infrastructure successfully managed massive traffic spikes post-I/O keynote with no API downtime.
Best for: High-speed agent workflows, general-purpose everyday use
Gemini Omni — new multimodal family
Launched: Google I/O 2026 | First model: Gemini Omni Flash
Introduced as Google's premier cross-modal creative model. The first rollout, Gemini Omni Flash, natively combines text, audio, images, and video inputs to generate and text-edit high-quality video outputs while maintaining perfect scene, character, and physics continuity. All output includes SynthID watermarking.
★ What's newGemini Omni Flash: native text + audio + image + video input, high-quality video output with scene/character/physics continuity. SynthID watermarking on all generated content. Available via Flow and YouTube Shorts.
Best for: Native video generation and editing, creative multimodal workflows
Gemini 3.5 Pro — coming next month
Status: final testing | Deployment: scheduled next month
Announced to be in final testing, with an official deployment scheduled for next month.
Best for: Watch this space — no action needed yet
Gemini Spark — autonomous cloud agent
Launched: Google I/O 2026 | Platform: Google Cloud VMs
An always-on, autonomous agent platform powered by Gemini 3.5 Flash. Running on Google Cloud VMs, Spark can review credit card statements for hidden subscriptions, track school updates from emails, compile notes into Docs, and safely perform multi-step actions across third-party apps like Instacart and OpenTable — requiring final user confirmation for purchases.
★ What's newGemini Spark: always-on autonomous background agent running on Google Cloud. Supports multi-step actions across third-party apps with user confirmation required for purchases. Continuous cloud automation without needing to stay in an active session.
Best for: Autonomous background tasks, continuous cloud automation
The 25-year Search upgrade
Launched: Google I/O 2026 | Rollout: concurrent with May 21 core search update
Google Search rolled out its biggest interface overhaul in over two decades, replacing the standard search box with an expanded AI Search Box. Users can input full natural language queries alongside images, video clips, and entire Chrome tabs simultaneously.
★ What's newAI Search Box replaces the classic Google search box. Accepts text, images, video clips, and full Chrome tabs as simultaneous inputs. May 21 core ranking algorithm update running concurrently.
Best for: Advanced web search, multimodal research queries
Gemini in Chrome — Android and Desktop
Android: coming next month | Desktop: available now
Coming to mobile next month, it introduces auto browse to automate digital chores (pulling event ticket details to book local parking). Desktop users gain Skills in Chrome, which saves complex multi-tab prompts into one-click reusable tools.
★ What's newAuto browse (mobile, next month): automates digital chores from within Chrome. Skills in Chrome (desktop): saves complex multi-tab prompts as reusable one-click tools.
Best for: Advanced web and coding agents, reusable browser automation
Android XR and Smart Glasses
Google teased an Android XR collaboration with Samsung, Gentle Monster, and Warby Parker, showcasing prototype smart glasses capable of real-world text translation, Gemini voice chatting, and real-time audio translation.
★ What's newPrototype smart glasses: real-world text translation, Gemini voice chat, real-time audio translation. Collaboration: Samsung, Gentle Monster, Warby Parker.
Best for: Watch this space — prototype stage, no release date announced
Plans and Pricing
AI subscription tier restructuring announced at Google I/O:
| Plan |
Price |
Key Includes |
| AI Ultra |
$99/mo |
5x higher usage limits vs. standard $20 Pro plan, 20TB storage, priority Antigravity developer tools |
| AI Ultra |
$200/mo |
20x higher usage limits, exclusive Project Genie access (interactive 3D from Street View) |
Antigravity 2.0 replaces legacy developer platforms and evolves into a full agentic ecosystem with a standalone desktop application and CLI. All Gemini CLI users are urged to migrate to Antigravity CLI immediately.
Microsoft Copilot
Dateline: May 22, 2026 | Next update: May 29, 2026
The biggest story of the week is model integration: Microsoft rolled out GPT-5.5 models across Copilot experiences, significantly improving reasoning, summarization, and writing quality.
GPT-5.5 models integrated into Copilot
Announced: May 19, 2026 | Effective: immediate | Applies to: Microsoft 365 Copilot (all tiers)
GPT-5.5 models now power Copilot, offering faster responses, deeper reasoning, and smarter writing assistance across all enterprise workflows.
★ What's newModels: GPT-5.5 Instant + GPT-5.5 Thinking now power all Copilot tiers. Improvements: reasoning quality, context understanding, summarization accuracy. Enterprise workflows benefiting: email drafting, meeting summaries, document creation, data analysis, presentations.
Technical detailsModels: GPT-5.5 Instant + GPT-5.5 Thinking | Applies to: all Microsoft 365 Copilot tiers | Effective: immediate from May 19, 2026
Best for: All Copilot users — especially enterprise teams needing higher-quality outputs
Outlook email grounding in Copilot Chat
Platform: Windows + Web | Availability: General
Users can now add emails or text from threads directly into Copilot Chat prompts, enabling context-aware answers without switching between apps.
★ What's newImplicit grounding in Outlook: insert email sections into Copilot Chat. Summarize, analyze, or extract action items from email content directly in the chat interface.
Best for: Business users managing large volumes of email
PDF opening inside Copilot Chat
Platform: Windows, Mac, Web | Availability: General
PDFs now open directly inside Copilot Chat, allowing summarization, highlighting, and Q&A without switching apps.
★ What's newPDFs open inline in Copilot — no need to leave the chat. Summarization and extraction supported. Reduces workflow interruptions for document-heavy tasks.
Best for: Researchers, analysts, legal teams, document-heavy workflows
App Launcher "Waffle" returns
Platform: Microsoft 365 | Availability: All plans
The classic "Waffle" app launcher is back, improving navigation across Outlook, Word, Excel, Teams, OneDrive, and PowerPoint.
★ What's newRestores the older launcher design by popular request. Quick access to all Microsoft 365 apps from a single consistent entry point.
Best for: Longtime Microsoft 365 users, productivity-focused teams
Researcher + Notebooks upgrades
Platform: Microsoft 365 Copilot | Availability: General
Researcher now provides deeper summaries and smarter document analysis. Copilot Notebooks gained new organisational features for managing research workflows.
★ What's newResearcher: improved information gathering and analysis. Notebooks: enhanced organisation and summarization. Now competitive with standalone AI research assistants.
Best for: Knowledge workers, students, enterprise research teams
Plans and Pricing
No pricing changes this week. Updates focus on feature expansion and model improvements. GPT-5.5 included in all Copilot tiers at no additional cost. Outlook grounding, PDF opening, Waffle, and Researcher upgrades all included in Microsoft 365 — no action needed.
Datum: 22. maj 2026. | Sledeće izdanje: 29. maj 2026.
Još jedna gusta nedelja sa tri glavne priče: Andrej Karpathy (suosnivač OpenAI-a, direktor Tesla AI-a) pridružio se Anthropic-ovom timu za pre-training 19. maja; samohostovani sandboxovi i MCP tuneli lansirani su za Claude Managed Agents na Code with Claude London događaju istog dana; a Gates fondacija — već najavljena 14. maja — detaljno je opisana kao Anthropic-ovo najveće partnerstvo usmereno na misiju ($200M tokom četiri godine). PwC je proširio Anthropic savez radi sertifikacije 30.000 profesionalaca na Claude-u, a API je dobio dijagnostiku keša u javnoj beti.
Claude / Anthropic
Andrej Karpathy se pridružuje Anthropic-u — tim za pre-training
Najavljeno: 19. maj 2026. | Uloga: Member of Technical Staff, pre-training | Izveštava: Nick Joseph
Andrej Karpathy — osnivački član OpenAI-a, bivši viši direktor Tesla AI-a i najpraćeniji AI edukator na svetu — pridružio se Anthropic-u 19. maja. Radiće pod vođom tima za pre-training Nickom Josephom sa specifičnim mandatom: pokrenuti novi tim koji koristi sam Claude za ubrzavanje istraživanja pre-traininga. Ovo je direktna opklada da AI-potpomognuto eksperimentisanje — pokretanje više ablacija, sažimanje internih rezultata, pisanje infrastrukturnog koda i predlaganje receptura za trening — jeste sledeća konkurentska poluga u trci za frontier modele.
★ Šta je novoKarpathy počinje u Anthropic-u 19. maja. Vodiće novu grupu unutar pre-traininga usmerenu na korišćenje Claude-a za ubrzavanje petlje istraživanja treninga. Pauzirao je svoju obrazovnu kompaniju Eureka Labs kako bi se pridružio. Bio je javno oduševljen Claude Code-om od januara 2026. Njegovo zapošljavanje dolazi dok godišnja stopa prihoda Anthropic-a dostiže $30B, a Claude Code sam generiše oko $2,5B godišnjeg prihoda.
Tehnički detaljiTim: pre-training | Vođa: Nick Joseph | Misija: AI-ubrzano istraživanje pre-traininga | Pozadina: osnivanje OpenAI-a (2015.), Tesla Autopilot (2017–2022.), OpenAI (2023.), Eureka Labs (2024–2026.) | Obrazovni rad: planira da nastavi kada dođe vreme
Najpogodnije za: Pratite ovaj prostor — Karpathyjev fokus na AI-ubrzano istraživanje pre-traininga je dugoročni signal, a ne neposredna promena proizvoda
Claude Managed Agents — samohostovani sandboxovi + MCP tuneli
Najavljeno: 19. maj 2026. na Code with Claude London | Samohostovani sandboxovi: javna beta | MCP tuneli: istraživački pregled (zahtevajte pristup)
Najavljeni na prvoj međunarodnoj konferenciji za programere Anthropic-a — Code with Claude London — samohostovani sandboxovi i MCP tuneli zajedno rešavaju ključnu prepreku za produkcijsko uvođenje enterprise agenata: podaci i kod koji napuštaju bezbednosni perimetar klijenta. Petlja orkestracije agenta ostaje na Anthropic-ovoj infrastrukturi; izvršavanje alata i pristup privatnim servisima prelaze unutar klijentovog perimetra.
★ Šta je novoSamohostovani sandboxovi (javna beta): izvršavanje alata sada radi na vašoj sopstvenoj infrastrukturi ili putem upravljanih provajdera — Cloudflare, Daytona, Modal ili Vercel. Vaše mrežne politike, revizijsko beleženje i bezbednosni alati primenjuju se po podrazumevanom podešavanju. MCP tuneli (istraživački pregled): lagan gateway koji postavite uspostavlja jednu izlaznu šifrovanu vezu — bez ulaznih pravila zaštitnog zida, bez javnih krajnjih tačaka. Agenti pristupaju internim bazama podataka, privatnim API-jima, sistemima za praćenje tiketa i bazama znanja kao MCP alatima bez izlaganja tih servisa javnom internetu. Takođe: MCP server i konfiguracije alata sada se mogu ažurirati na aktivnoj sesiji (bez ponovnog pokretanja); veliki izlazi alata iznad 100K tokena automatski se prebacuju u sandbox fajl.
Tehnički detaljiSamohostovani sandboxovi: javna beta | Podržani provajderi: Cloudflare (microVM), Daytona, Modal, Vercel | MCP tuneli: istraživački pregled, zahtevajte pristup putem Claude Console podešavanja radnog prostora | Orkestracija ostaje na Anthropic infrastrukturi | Ažuriranja MCP konfiguracije uživo: da, bez ponovnog pokretanja sesije | Spill velikih izlaza: >100K tokena auto-prebacivanje u sandbox fajl | Oba pod managed-agents-2026-04-01 zaglavljem
Najpogodnije za: Preduzeća u regulisanim industrijama (finansije, zdravstvo, pravo, vlada) koja zahtevaju izvršavanje agenata unutar sopstvenog bezbednosnog perimetra
Gates fondacija — partnerstvo od $200M na četiri godine
Najavljeno: 14. maj 2026. | Obaveza: $200M tokom četiri godine | Fokus: globalno zdravstvo, obrazovanje, ekonomska mobilnost
Anthropic se obavezao na 200 miliona dolara tokom četiri godine — u grant finansiranju, Claude kreditima za korišćenje i ugrađenoj inženjerskoj podršci — za partnerstvo sa Fondacijom Bila i Melinde Gejts. To je najveća AI obaveza Gates fondacije do danas (4x od $50M koje je dala OpenAI-u u januaru za uži program afričkih klinika). Radom rukovodi Anthropic-ov tim za Beneficial Deployments.
★ Šta je novoZdravstveni fokus: ubrzavanje skrininga kandidata za vakcine i lekove za zanemarene bolesti (uključujući polio, HPV, preeklampsiju) i poboljšanje prognoza za uvođenje tretmana malarije i tuberkuloze. Partnerstvo će služiti oko 4,6 milijardi ljudi u zemljama sa niskim i srednjim prihodima. Obrazovni fokus: alati za poboljšanje ishoda od K-12 u SAD-u, subsaharskoj Africi i Indiji. Deliverables javnih dobara uključuju otvorene zdravstvene skupove podataka, evaluacione benčmarkove i grafove znanja.
Tehnički detaljiObaveza: $200M tokom 4 godine | Komponente: grant finansiranje + Claude krediti za korišćenje + Anthropic inženjerska podrška | Vodi: Anthropic Beneficial Deployments tim | Partner: Fondacija Bila i Melinde Gejts | Javna dobra: otvoreni skupovi podataka, benčmarkovi, grafovi znanja objavljeni za širu istraživačku zajednicu
Najpogodnije za: Istraživači globalnog zdravlja, edukatori, nevladine organizacije i kreatori politika — informativno za komercijalne korisnike
PwC — proširena strateška alijansa
Najavljeno: 14. maj 2026. | Obim: globalna PwC radna snaga
Anthropic i PwC objavili su veliku ekspanziju svoje strateške alijanse. Uvođenje počinje sa PwC-ovim timovima u SAD-u i širi se na stotine hiljada profesionalaca globalno, uz zajednički Centar izvrsnosti i program sertifikacije koji cilja 30.000 PwC profesionalaca obučenih i sertifikovanih na Claude-u.
★ Šta je novoClaude Code i Claude Cowork uvode se u celokupnu globalnu radnu snagu PwC-a. Uspostavljen zajednički Anthropic–PwC Centar izvrsnosti. 30.000 PwC profesionalaca će biti obučeno i sertifikovano na Claude-u. Uvođenje počinje sa timovima u SAD-u, širi se globalno. Ovo PwC čini jednim od najvećih pojedinačnih enterprise uvođenja Claude-a do danas.
Tehnički detaljiProizvodi: Claude Code + Claude Cowork | Početna geografija: SAD, globalno širenje | Centar izvrsnosti: zajednički Anthropic–PwC | Ciljana sertifikacija: 30.000 profesionalaca
Najpogodnije za: Enterprise kupci koji procenjuju uvođenje u profesionalne usluge velikih razmera; konsultantske i savetodavne firme
API — dijagnostika keša u javnoj beti
Platforma: Claude API | Beta zaglavlje: cache-diagnosis-2026-04-07 | Dostupnost: svi API korisnici
Keš upita je jedan od najefikasnijih alata za smanjenje troškova na Claude API-ju, ali do ove nedelje nije postojao način da se zna zašto je došlo do promašaja keša. Dijagnostika keša, sada u javnoj beti, precizno govori gde je prefiks keša upita odstupio od prethodnog okreta — omogućavajući debagovanje i ispravku problema keširanja bez nagađanja.
★ Šta je novoProsledite diagnostics.previous_message_id u Messages zahtevu i API vraća polje cache_miss_reason koje objašnjava gde je prefiks keša odstupio. Omogućava sistematsko debagovanje promašaja keša upita u višeokretnim agentskim procesima.
Tehnički detaljiBeta zaglavlje: cache-diagnosis-2026-04-07 | Parametar: diagnostics.previous_message_id | Polje odgovora: cache_miss_reason | Radi na: Messages API | Slučaj upotrebe: debagovanje promašaja keša upita u agentskim i višeokretnim procesima
Najpogodnije za: API programeri koji optimizuju troškove i kašnjenje kroz keš upita
Claude Code — Agent Teams, ispravke dodataka, alati za sesije
Platforma: terminal / VS Code / veb / mobilni | Dostupnost: svi planovi
Claude Code je objavio široko ažuriranje usmereno na stabilnost Agent Teams, poboljšanja dodataka i poliranje alata za sesije. Agent Teams — kolaborativne višeagentske sesije gde imenovani saradnici zajedno rade na zadatku — primili su nekoliko ispravki pouzdanosti nakon što su non-ASCII imena uzrokovala greške API poziva i rušila sesije.
★ Šta je novoIspravke Agent Teams: saradnici sa non-ASCII imenima više ne greše u svakom API pozivu zbog nevalidnog kodiranja zaglavlja. Marketplace dodataka: dodavanje/ažuriranje sada poštuje CLAUDE_CODE_PLUGIN_PREFER_HTTPS; /plugin se vraća na listu Installed nakon enable/disable/uninstall akcija. /doctor sada prikazuje primer exec-forme kada hook komandi nedostaje polje command. Oporavak od zastoja pre-response streama poboljšan — sada pokušava streaming jednom pre nego što pređe na sporiji non-streaming put. SDK/headless MCP pokretanje je do 2 sekunde brže sa sporim MCP serverima. Ispravka: beskonačna petlja gde veština koja koristi context: fork ponavljano poziva samu sebe. Ispravka: /review koji koristi zastareli projectCards GraphQL upit.
Tehnički detaljiAgent Teams: ispravka kodiranja zaglavlja za non-ASCII imena | CLAUDE_CODE_PLUGIN_PREFER_HTTPS: sada se poštuje pri dodavanju/ažuriranju na marketplaceu | MCP headless pokretanje: do 2s brže | Zastoj streama: pokušaj streaming-jednom pre prelaza | Ispravka: beskonačna petlja context: fork | Ispravka: zastareli GraphQL projectCards | Ispravka: zastareli baner 'Failed to install Anthropic marketplace' | Ispravka: PR bedž se ne ažurira nakon gh pr create u sesiji
Najpogodnije za: Programeri koji koriste Agent Teams, MCP servere u headless SDK modu, marketplace dodataka
Limiti korišćenja — zabeleženo zaoštravanje za Pro i Max
Prijavljeno: 14. maj 2026. | Utiče na: Pro i Max potrošačke planove
Anthropic je tiho zaostrio limite korišćenja Claude-a za Pro i Max pretplatnike. API nije bio pogođen. Ovo je promena samo za potrošačke planove i stigla je ubrzo nakon udvostručavanja Claude Code petočasovnih limita putem SpaceX compute dogovora najavljenog 6. maja. Anthropic nije objavio detaljan pregled novih limita.
★ Šta je novoLimiti korišćenja Pro i Max plana zaoštreni od oko 14. maja. API cene i limiti nepromenjeni. Neto efekat na Claude Code petočasovne limite u odnosu na udvostručavanje od 6. maja je nejasan — pratite svoju kontrolnu tablu korišćenja ako ste intenzivni Pro ili Max korisnik.
Tehnički detaljiPogođeni: Pro ($20/mes.) i Max ($100–200/mes.) potrošački planovi | API: nije pogođen | Izvor: Axios, 14. maj 2026. | Nema zvaničnog Anthropic saopštenja o specifičnostima | Praktičan uticaj: najvidljivije na dugim Claude Code sesijama bez automatskog kompaktiranja konteksta
Najpogodnije za: Intenzivni Pro i Max pretplatnici — proverite kontrolnu tablu korišćenja; nema potrebe za akcijom za API korisnike
Planovi i cene
Nema promena API cena ove nedelje. Zaoštravanje limita korišćenja na Pro i Max planovima je jedina promena okrenuta ka potrošačima i nije zvanično kvantifikovana. Potrošnja Claude Platforme na AWS-u putem AWS Marketplace ostaje ista kao i nativne Claude API stope.
Tehnički detaljiOpus 4.7: $5/$25 po MTok | Sonnet 4.6: $3/$15 po MTok | Haiku 4.5: niskobudžetni nivo | Claude Platform na AWS-u: potrošnja putem AWS Marketplace | Gates fondacija i PwC: enterprise/misijom usmerene cene, kontaktirajte Anthropic | Samohostovani sandboxovi + MCP tuneli: uključeni u Managed Agents (beta)
ChatGPT / OpenAI
Datum: 22. maj 2026. | Sledeće izdanje: 29. maj 2026.
Tokom protekle nedelje, OpenAI je nastavio da poboljšava pouzdanost agenata, multimodalne procese i enterprise integracije, tiho usavršavajući sistemsku konzistentnost u profesionalnim slučajevima upotrebe.
GPT-5.3 Standard — podrazumevani model
Objavljivanje: kasno 2025. | Cena: uključeno | Dostupnost: svi korisnici
Pouzdanija obrada dugih razgovora i razgovora mešovitog formata.
★ Šta je novoPoboljšana konzistentnost pri prebacivanju između zadataka zasnovanih na tekstu, dokumentima i slikama.
Tehnički detaljiKontekst ~128k | Izlaz ~4k–8k | Poboljšano upravljanje multimodalnim kontekstom
Najpogodnije za: Opšta upotreba
GPT-5.3 Pro — model visokog rezonovanja
Objavljivanje: kasno 2025. | Cena: Pro | Dostupnost: Pro/Enterprise
Stabilniji rad na slojevitim analitičkim zadacima.
★ Šta je novoPoboljšan kontinuitet rezonovanja u veoma dugim izlazima.
Tehnički detaljiKontekst ~200k (procena) | Smanjena degradacija kroz proširene lance rezonovanja
Najpogodnije za: Duboka analiza
GPT-5.3 Mini — rezervni model
Objavljivanje: kasno 2025. | Cena: niskobudžetno | Dostupnost: svi
Brži i glađi lagani odgovori.
★ Šta je novoPoboljšana preciznost rutiranja između Mini i Standard modela.
Tehnički detaljiKontekst ~64k | Bolja dinamična alokacija zaključivanja
Najpogodnije za: Brzi zadaci
Agent Mode
Rukuje višekoračnim procesima.
★ Šta je novoPoboljšana pouzdanost u izvršavanju strukturiranih procesa tokom dužih sesija.
Tehnički detaljiBolja persistencija zadataka | Smanjene stope prekida procesa
Najpogodnije za: Delegiranje zadataka
Deep Research
Kombinuje pregledanje interneta i rezonovanje.
★ Šta je novoPoboljšena obrada konfliktnih informacija iz više izvora.
Tehnički detaljiPoboljšan kanal sinteze i ponderisanja izvora
Najpogodnije za: Istraživanje
Memorija i Projekti
Trajni kontekst kroz razgovore.
★ Šta je novoBolje davanje prioriteta kontekstu aktivnog projekta nad starijom konverzacijskom memorijom.
Tehnički detaljiPoboljšano filtriranje relevantnosti | Efikasnije preuzimanje memorije
Najpogodnije za: Tekući procesi
Napredni glasovni mod
★ Šta je novoPrirodniji prelazi između konverzacijskih okreta.
Tehnički detaljiPoboljšeno kašnjenje preuzimanja okreta | Bolje konverzacijsko tempo
Najpogodnije za: Glasovna interakcija
ChatGPT za kliničare
★ Šta je novoPoboljšana struktura i čitljivost generisanih kliničkih sažetaka i administrativnih izlaza.
Tehnički detaljiDalje domainsko podešavanje | Pojačane medicinske bezbednosne zaštite
Najpogodnije za: Klinička podrška (nedijagnostička asistencija)
Enterprise i procesne integracije
★ Šta je novoPoboljšana pouzdanost u integracijama koje uključuju dokumente, kolaborativne procese i trajne projekte. Bolja orkestracija kroz povezane alate i memorijske sisteme.
Najpogodnije za: Organizaciono uvođenje
Planovi i cene
Nema značajnih promena ove nedelje. Cene stabilne | Struktura API-ja nepromenjena.
Gemini (Google)
Datum: 22. maj 2026. | Sledeće izdanje: 29. maj 2026.
Ova nedelja označila je početak Google I/O 2026, zvanično premeštajući ekosistem u "Agentsku Gemini eru." Google je predstavio potpuno novu generaciju modela, multimodalne video arhitekture i autonomne cloud agente.
Gemini 3.5 Flash — novi podrazumevani model
Lansiran: Google I/O 2026. | Dostupnost: Gemini app i Google Search (podrazumevano)
Lansiran kao podrazumevani mehanizam u Gemini appu i Google Pretrazi. Radi četiri puta brže od ostalih frontier modela u izlaznim tokenima po sekundi i nadmašuje Gemini 3.1 Pro u ključnim benčmarkovima logike i kodiranja.
★ Šta je novoGemini 3.5 Flash je sada podrazumevani model u Gemini appu i Google Pretrazi. 4x brži propusni kapacitet izlaznih tokena od konkurentskih frontier modela. Nadmašuje Gemini 3.1 Pro na benčmarkovima logike i kodiranja. Infrastruktura uspešno upravljala masovnim skokovima saobraćaja nakon I/O keynote-a bez API prekida.
Najpogodnije za: Brzinski agentski procesi, opšta svakodnevna upotreba
Gemini Omni — nova multimodalna porodica
Lansiran: Google I/O 2026. | Prvi model: Gemini Omni Flash
Predstavljen kao Google-ov vodeći cross-modalni kreativni model. Prvo izdanje, Gemini Omni Flash, izvorno kombinuje unose teksta, zvuka, slika i video zapisa za generisanje i tekstualno uređivanje visokokvalitetnih video izlaza uz održavanje savršenog kontinuiteta scene, likova i fizike. Svi izlazi uključuju SynthID vodeni žig.
★ Šta je novoGemini Omni Flash: nativni unos teksta + zvuka + slike + videa, visokokvalitetan video izlaz sa kontinuitetom scene/likova/fizike. SynthID vodeni žig na svim generisanim sadržajima. Dostupno putem Flow i YouTube Shorts.
Najpogodnije za: Nativno generisanje i uređivanje video zapisa, kreativni multimodalni procesi
Gemini 3.5 Pro — dolazi sledećeg meseca
Status: finalno testiranje | Uvođenje: planirano sledećeg meseca
Najavljeno je da je u finalnom testiranju, sa zvaničnim uvođenjem zakazanim za sledeći mesec.
Najpogodnije za: Pratite ovaj prostor — za sada nema potrebe za akcijom
Gemini Spark — autonomni cloud agent
Lansiran: Google I/O 2026. | Platforma: Google Cloud VM-ovi
Uvek aktivan, autonomni agentski platform koji pokreće Gemini 3.5 Flash. Pokrenut na Google Cloud VM-ovima, Spark može pregledati izvode kreditnih kartica za skrivene pretplate, pratiti školska obaveštenja iz e-pošte, kompajlirati beleške u Docs i bezbedno izvoditi višekoračne akcije u aplikacijama trećih lica poput Instacart i OpenTable — uz obaveznu finalnu potvrdu korisnika za kupovine.
★ Šta je novoGemini Spark: uvek aktivan autonomni pozadinski agent koji radi na Google Cloud-u. Podržava višekoračne akcije u aplikacijama trećih lica uz obaveznu potvrdu korisnika za kupovine. Kontinuirana cloud automatizacija bez potrebe za aktivnom sesijom.
Najpogodnije za: Autonomni pozadinski zadaci, kontinuirana cloud automatizacija
Nadogradnja pretrage posle 25 godina
Lansiran: Google I/O 2026. | Uvođenje: paralelno sa ažuriranjem jezgra pretrage od 21. maja
Google Pretraga je uvela najveću promenu interfejsa u više od dve decenije, zamenjujući standardni okvir za pretragu proširenim AI okvirom za pretragu. Korisnici mogu istovremeno unositi upite na prirodnom jeziku zajedno sa slikama, video isečcima i celim Chrome karticama.
★ Šta je novoAI okvir za pretragu zamenjuje klasičan Google okvir za pretragu. Prihvata tekst, slike, video isečke i pune Chrome kartice kao istovremene unose. Ažuriranje osnovnog algoritma rangiranja od 21. maja radi paralelno.
Najpogodnije za: Napredna veb pretraga, multimodalni istraživački upiti
Gemini u Chrome-u — Android i Desktop
Android: dolazi sledećeg meseca | Desktop: dostupno sada
Stiže na mobilne uređaje sledećeg meseca sa auto browse funkcijom za automatizaciju digitalnih zadataka (prikupljanje detalja ulaznica za događaje radi rezervacije lokalnog parkinga). Desktop korisnici dobijaju Veštine u Chrome-u, koji čuva složene upite za više kartica kao alate koji se ponovo mogu koristiti jednim klikom.
★ Šta je novoAuto browse (mobilni, sledećeg meseca): automatizuje digitalne zadatke iz Chrome-a. Veštine u Chrome-u (desktop): čuva složene upite za više kartica kao alate jednim klikom koji se mogu ponovo koristiti.
Najpogodnije za: Napredni veb i kodirački agenti, automatizacija pregledača za višekratnu upotrebu
Android XR i pametne naočare
Google je nagovestio saradnju na Android XR-u sa Samsung-om, Gentle Monster-om i Warby Parker-om, prikazujući prototipove pametnih naočara sposobnih za prevod teksta u stvarnom svetu, glasovni razgovor sa Gemini-jem i prevod zvuka u realnom vremenu.
★ Šta je novoPrototipovi pametnih naočara: prevod teksta u stvarnom svetu, glasovni razgovor sa Gemini-jem, prevod zvuka u realnom vremenu. Saradnja: Samsung, Gentle Monster, Warby Parker.
Najpogodnije za: Pratite ovaj prostor — u fazi prototipa, bez najavljenog datuma objavljivanja
Planovi i cene
Restrukturiranje AI pretplatnih nivoa najavljeno na Google I/O:
| Plan |
Cena |
Šta uključuje |
| AI Ultra |
$99/mes. |
5x viši limiti korišćenja vs. standardni $20 Pro plan, 20TB skladišta, prioritetni Antigravity developer alati |
| AI Ultra |
$200/mes. |
20x viši limiti korišćenja, ekskluzivni pristup Project Genie-u (interaktivni 3D iz Street View podataka) |
Antigravity 2.0 zamenjuje nasleđene developer platforme i razvija se u potpun agentski ekosistem sa standalone desktop aplikacijom i CLI-jem. Svi Gemini CLI korisnici se pozivaju da odmah migriraju na Antigravity CLI.
Microsoft Copilot
Datum: 22. maj 2026. | Sledeće izdanje: 29. maj 2026.
Najveća priča nedelje je integracija modela: Microsoft je uveo GPT-5.5 modele u Copilot iskustva, značajno poboljšavajući rezonovanje, sažimanje i kvalitet pisanja.
GPT-5.5 modeli integrisani u Copilot
Najavljeno: 19. maj 2026. | Važi od: odmah | Primenjuje se na: Microsoft 365 Copilot (svi nivoi)
GPT-5.5 modeli sada pokreću Copilot, nudeći brže odgovore, dublje rezonovanje i pametniju asistenciju pri pisanju u svim enterprise procesima.
★ Šta je novoModeli: GPT-5.5 Instant + GPT-5.5 Thinking sada pokreću sve Copilot nivoe. Poboljšanja: kvalitet rezonovanja, razumevanje konteksta, tačnost sažimanja. Enterprise procesi koji imaju koristi: pisanje e-pošte, sažimanje sastanaka, kreiranje dokumenata, analiza podataka, prezentacije.
Tehnički detaljiModeli: GPT-5.5 Instant + GPT-5.5 Thinking | Primenjuje se na: sve Microsoft 365 Copilot nivoe | Važi od: odmah od 19. maja 2026.
Najpogodnije za: Svi Copilot korisnici — posebno enterprise timovi koji trebaju kvalitetnije izlaze
Uzemljenje e-pošte Outlook-a u Copilot Chat-u
Platforma: Windows + Veb | Dostupnost: Opšta dostupnost
Korisnici sada mogu dodavati e-poruke ili tekst iz niti direktno u Copilot Chat upite, omogućavajući odgovore svesne konteksta bez prebacivanja između aplikacija.
★ Šta je novoImplicitno uzemljenje u Outlook-u: ubacite delove e-pošte u Copilot Chat. Sažimajte, analizujte ili izvlačite stavke akcija iz sadržaja e-pošte direktno u interfejsu četa.
Najpogodnije za: Poslovni korisnici koji upravljaju velikim volumenom e-pošte
Otvaranje PDF-ova unutar Copilot Chat-a
Platforma: Windows, Mac, Veb | Dostupnost: Opšta dostupnost
PDF-ovi se sada otvaraju direktno unutar Copilot Chat-a, omogućavajući sažimanje, isticanje i Q&A bez prebacivanja aplikacija.
★ Šta je novoPDF-ovi se otvaraju inline u Copilot-u — bez napuštanja četa. Podržano sažimanje i ekstrakcija. Smanjuje prekide procesa za zadatke s mnogo dokumenata.
Najpogodnije za: Istraživači, analitičari, pravni timovi, procesi s mnogo dokumenata
Pokretač aplikacija "Waffle" se vraća
Platforma: Microsoft 365 | Dostupnost: Svi planovi
Klasični pokretač aplikacija "Waffle" se vratio, poboljšavajući navigaciju u Outlook-u, Word-u, Excel-u, Teams-u, OneDrive-u i PowerPoint-u.
★ Šta je novoVraća stariji dizajn pokretača na zahtev korisnika. Brz pristup svim Microsoft 365 aplikacijama sa jedne konzistentne ulazne tačke.
Najpogodnije za: Dugogodišnji Microsoft 365 korisnici, timovi fokusirani na produktivnost
Nadogradnje Researcher-a i Notebooks-a
Platforma: Microsoft 365 Copilot | Dostupnost: Opšta dostupnost
Researcher sada pruža dublje sažetke i pametniju analizu dokumenata. Copilot Notebooks su dobili nove organizacione funkcionalnosti za upravljanje istraživačkim procesima.
★ Šta je novoResearcher: poboljšano prikupljanje informacija i analiza. Notebooks: poboljšana organizacija i sažimanje. Sada konkurentno sa standalone AI istraživačkim asistentima.
Najpogodnije za: Radnici znanja, studenti, enterprise istraživački timovi
Planovi i cene
Nema promena cena ove nedelje. Ažuriranja se fokusiraju na proširenje funkcionalnosti i poboljšanja modela. GPT-5.5 uključen u sve Copilot nivoe bez dodatnih troškova. Uzemljenje Outlook-a, otvaranje PDF-ova, Waffle i nadogradnje Researcher-a uključeni u Microsoft 365 — nema potrebe za akcijom.
