# AI Advancements Atlas

> Original-source dated AI milestones from 2025–2026; a separate 2027+ questions watchlist and searchable practical use cases.

Canonical: https://ainextia.com/advancements/
Publisher: https://ainextia.com/

## Dated milestones and annual 2025 observations
- 2026-11-05 | Specialized AI assists authorized cybersecurity work | Safety & Security | OpenAI | Evidence: Product capability | Metric:  | GPT-6-Cyber was released for authorized vulnerability research and remediation verification. | Possible: Authorized security teams can test controlled defensive investigation and fix verification. | Limit: Restricted access and authorized environments are required; not consumer-facing unrestricted use. | Source: https://developers.openai.com/api/docs/changelog
- 2026-11-04 | AI participates in a multi-step biology discovery workflow | Science | Anthropic | Evidence: Research demonstration | Metric:  | Anthropic documented an AI-assisted workflow for studying vault-like particles using biological tools and cryo-EM. | Possible: Explore research assistance across literature, protocol drafting and image analysis with scientists in control. | Limit: A single case study does not establish universal scientific accuracy. | Source: https://www.anthropic.com/research/investigating-vault-particles-with-claude
- 2026-11-04 | Animated GIF generation arrives in image workflows | Creative AI | OpenAI | Evidence: Product capability | Metric:  | Images 2.6 added animated GIF creation, stronger iterative editing and image controls. | Possible: Create animated concepts, brand mascots and lightweight demonstrations with review. | Limit: Generated animations may contain temporal or factual errors; usage rights still matter. | Source: https://openai.com/index/introducing-images-2-6/
- 2026-11-03 | AI-agent chemistry demonstration spans compound search and design | Biology & Medicine | Google DeepMind | Evidence: Research demonstration | Metric: One-hour demonstrated cheminformatics workflow | DeepMind showed an agent comparing compounds and proposing molecule structures and synthesis routes. | Possible: Drug researchers can streamline literature and chemical-data exploration. | Limit: Proposed candidates are not synthesized or proven medicines. | Source: https://deepmind.google/blog/agentic-cheminformatics-with-google-antigravity/
- 2026-11-02 | Research explores cryptographic proof of model inference | Infrastructure | Anthropic | Evidence: Research demonstration | Metric:  | Research on provable inference explores proving which model served a result without revealing model weights. | Possible: Follow potential provenance controls for enterprise AI deployments. | Limit: Not an ordinary production verification feature; overhead and deployment remain open questions. | Source: https://www.anthropic.com/research/provable-inference
- 2026-11 | Agentic cheminformatics explores scientific tooling | Science | Google DeepMind | Evidence: Research demonstration | Metric:  | DeepMind listed November research on agentic chemistry information processing. | Possible: Watch how scientific agents connect literature, molecular data and validated tools. | Limit: Headline-level evidence only; research claims require paper-level review. | Source: https://deepmind.google/blog/
- 2026-11 | Gemini Omni 1.2 Flash expands multimedia model options | Models | Google DeepMind | Evidence: Product capability | Metric:  | DeepMind listed a November update focused on faster multimedia generation and longer clips. | Possible: Compare supported multimedia tasks and costs on your actual workflow. | Limit: Month-level announcement; exact supported features require current docs. | Source: https://deepmind.google/blog/
- 2026-10-26 | GPT-6.1 Terra extends computer use at scale | Agents | OpenAI | Evidence: Product capability | Metric:  | OpenAI's release archive describes Terra as a model focused on computer-use tasks. | Possible: Benchmark real browser and desktop actions with explicit permission gates. | Limit: Model availability and task reliability depend on access and environment. | Source: https://openai.com/research/index/release/
- 2026-10-15 | AI research agents prioritize antimicrobial peptides | Biology & Medicine | Google DeepMind | Evidence: Research demonstration | Metric: Lab validation pending | DeepMind described an agent identifying a plausible antimicrobial peptide through protein data searches and structure analysis. | Possible: Researchers can narrow a huge search space before expensive experiments. | Limit: Predicted biological activity is not experimentally demonstrated efficacy. | Source: https://deepmind.google/blog/accelerating-antimicrobial-discovery/
- 2026-10-05 | PIE predicts cell responses in unseen conditions | Biology & Medicine | Arc Institute | Evidence: Research demonstration | Metric: Out-of-context perturbation generalization | Arc's PIE model uses biological priors to improve prediction of population-level cell responses. | Possible: Scientists can prioritize experiments in previously unseen settings. | Limit: Predictions need withheld-data testing and experimental replication. | Source: https://arcinstitute.org/news/pie
- 2026-10 | Genie 4 advances interactive world simulation | Creative AI | Google DeepMind | Evidence: Research demonstration | Metric:  | Google DeepMind announced a new generation of world-simulation research. | Possible: Watch for simulation-based prototyping and interactive scene exploration. | Limit: World simulations are not exact models of the physical world. | Source: https://deepmind.google/blog/
- 2026-09-08 | AlphaGenome Atlas maps 9 billion possible DNA variants | Science | Google DeepMind | Evidence: Research demonstration | Metric:  | The Atlas published predicted molecular effects for roughly nine billion possible single-letter variants. | Possible: Explore questions about gene regulation and candidate variants in research settings. | Limit: Predictions are computational, not validated clinical conclusions. | Source: https://deepmind.google/blog/alphagenome-atlas-a-predictive-map-of-every-possible-dna-letter-change-in-the-human-genome/
- 2026-09-08 | Nine billion DNA changes receive predictive annotations | Biology & Medicine | Google DeepMind | Evidence: Research demonstration | Metric: Nine billion possible single-letter variants | DeepMind released AlphaGenome Atlas predictions at human-genome scale. | Possible: Researchers can search a broad set of candidate molecular effects. | Limit: These are modeled changes rather than billions of laboratory observations. | Source: https://deepmind.google/blog/alphagenome-atlas-a-predictive-map-of-every-possible-dna-letter-change-in-the-human-genome/
- 2026-09 | Agentic video understanding enters model workflows | Creative AI | Google DeepMind | Evidence: Product capability | Metric:  | DeepMind announced agentic video understanding capabilities among its September model updates. | Possible: Explore video search, scene analysis and clip retrieval subject to permissions. | Limit: Feature availability and accuracy depend on provider and implementation. | Source: https://deepmind.google/blog/
- 2026-09 | Robotaxis accumulate hundreds of millions of driverless miles | Mobility & Transport | Waymo | Evidence: Real-world deployment | Metric: 271.3 million rider-only miles through June 2026 | Waymo published city-level mileage and safety-comparison methodology. | Possible: Large driving datasets support empirical safety evaluation rather than relying on demo clips. | Limit: Reported safety comparisons depend on exposure, location and benchmark definitions. | Source: https://waymo.com/safety/impact/
- 2026-09 | WeatherNext 3 expands AI weather-model research | Science | Google DeepMind | Evidence: Research demonstration | Metric:  | DeepMind announced a new generation of WeatherNext global weather forecasting models. | Possible: Investigate research-led weather modeling and uncertainty-aware forecast workflows. | Limit: Forecast performance depends on region, horizon and evaluation; never replace official warnings. | Source: https://deepmind.google/blog/
- 2026-08-18 | Flash-flood prediction expands from public disaster records | Climate & Weather | Google | Evidence: Research demonstration | Metric: 150-country flood forecasting coverage | Google documented wider flood forecasting and Groundsource methodology for extracting flash-flood events. | Possible: Researchers can build better geographically varied flood-risk baselines. | Limit: Model predictions are not official evacuation instructions. | Source: https://blog.google/innovation-and-ai/technology/research/flood-prediction-ai/
- 2026-07-09 | ChatGPT Work supports longer multi-app tasks | Agents | OpenAI | Evidence: Product capability | Metric:  | ChatGPT Work introduced agents that operate across web, files and apps to complete more substantial projects. | Possible: Test multi-source research, document production and cross-tool business processes. | Limit: Control permissions, verify outputs and check rollout availability. | Source: https://openai.com/index/chatgpt-for-your-most-ambitious-work/
- 2026-06-16 | Coding-agent use expands beyond software repairs | Software & Work | Anthropic | Evidence: Measured trend | Metric: Around 400,000 sessions analyzed | Anthropic documented coding agents increasingly supporting operations, writing and analysis tasks. | Possible: Organizations can evaluate expanding use cases across teams. | Limit: Usage growth is not controlled proof of productivity. | Source: https://www.anthropic.com/research/claude-code-expertise
- 2026-06-16 | Large-scale study examines real coding-agent usage | Coding & Agents | Anthropic | Evidence: Research demonstration | Metric:  | Anthropic studied about 400,000 Claude Code sessions to examine human-agent collaboration and expertise. | Possible: Use the study to design code-agent oversight and human review roles. | Limit: Session-based observational findings do not guarantee task success in another workflow. | Source: https://www.anthropic.com/research/claude-code-expertise
- 2026-06-09 | Live speech translation spans 70+ input languages | Accessibility & Language | Google | Evidence: Product capability | Metric: 70+ spoken languages supported | Google described continuous low-latency translation while preserving aspects of voice expression. | Possible: More natural multilingual interactions can be prototyped in products. | Limit: Noise, accents, mistakes and privacy remain significant concerns. | Source: https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-live-3-5-translate/
- 2026-05-27 | Researchers experiment widely with AI but fewer use agents | Education & Learning | Anthropic | Evidence: Measured trend | Metric: 81% chatbots vs 20% coding agents | A survey of 1,260 social scientists found substantial chatbot use but lower agent use. | Possible: Academic teams can target realistic training and workflow support. | Limit: Self-reported uptake does not show causally improved research. | Source: https://www.anthropic.com/research/coding-agents-social-sciences
- 2026-05-07 | AI-designed algorithms improve live infrastructure | Infrastructure & Energy | Google DeepMind | Evidence: Real-world deployment | Metric: 20% lower write amplification in Google Spanner example | Google described AlphaEvolve ideas integrated into TPU hardware and database heuristics. | Possible: Algorithm search can measurably improve computing systems in production. | Limit: Vendor-specific gains need separate testing before generalizing. | Source: https://deepmind.google/blog/alphaevolve-impact/
- 2026-05-07 | AlphaEvolve designs and improves algorithms across fields | Science | Google DeepMind | Evidence: Research demonstration | Metric:  | AlphaEvolve research described algorithm improvements and deployments in some Google infrastructure. | Possible: Explore algorithm-search assistance for narrowly testable optimization problems. | Limit: Company case studies do not prove equivalent gains for every organization. | Source: https://deepmind.google/blog/alphaevolve-impact/
- 2026-05-07 | Real-time voice models combine reasoning and translation | Accessibility & Language | OpenAI | Evidence: Product capability | Metric: Three specialized audio models | OpenAI introduced new streaming voice reasoning, translation and transcription models. | Possible: Developers can build complex voice interfaces with lower delay. | Limit: Test speech accuracy, consent and costs before deployment. | Source: https://openai.com/index/advancing-voice-intelligence-with-new-models-in-the-api/
- 2026-05 | AI materials models generate candidate inorganic structures | Materials & Manufacturing | Microsoft Research | Evidence: Research demonstration | Metric: Generative materials design | Microsoft Research documented MatterGen and MatterSim approaches for inverse materials design and simulation. | Possible: Researchers can rank property-constrained candidate crystals faster. | Limit: Candidates require synthesis and physical validation. | Source: https://www.microsoft.com/en-us/research/project/materials/
- 2026-03-05 | Early labor research finds uneven AI displacement signals | Economy & Labor | Anthropic | Evidence: Measured trend | Metric: No systematic unemployment rise detected | Anthropic compared observed AI usage with occupation exposure and labor data. | Possible: Researchers can monitor which tasks and jobs show evidence of disruption. | Limit: No causal claim of job destruction or permanent safety is supported. | Source: https://www.anthropic.com/research/labor-market-impacts
- 2026-03-04 | Evo 2 DNA foundation model reaches Nature | Biology & Medicine | Arc Institute | Evidence: Independent evidence | Metric: Open genomics model | Arc reported Evo 2's publication in Nature after its earlier preprint. | Possible: Open biological model research gains peer-reviewed publication and new application studies. | Limit: Not equivalent to a medically approved test. | Source: https://arcinstitute.org/news/evo-2-one-year-later
- 2026-02-11 | AI contributes to original mathematical research | Mathematics & Algorithms | Google DeepMind | Evidence: Research demonstration | Metric: Multiple worked open questions and papers | DeepMind described Aletheia work on mathematical conjectures with explicit human versus AI contribution levels. | Possible: Researchers can test AI as a proof assistant and conjecture partner. | Limit: The authors did not claim the highest-tier landmark advances. | Source: https://deepmind.google/blog/accelerating-mathematical-and-scientific-discovery-with-gemini-deep-think/
- 2026-02-10 | Deep research gains connected and bounded sources | Research | OpenAI | Evidence: Measured trend | Metric:  | Deep research added support for connected sources and trusted-site search restrictions. | Possible: Run scoped research across approved internal records and nominated sources. | Limit: Access varies by plan and connected source permissions. | Source: https://openai.com/index/introducing-deep-research/
- 2026-02-05 | Codex takes on longer professional computer tasks | Coding & Agents | OpenAI | Evidence: Product capability | Metric:  | GPT-5.3-Codex extended agentic coding into longer research, tool use and execution tasks. | Possible: Try multi-step code maintenance and review checkpoints rather than single-turn snippets. | Limit: Model-provider claims require representative testing and can change with newer releases. | Source: https://openai.com/index/introducing-gpt-5-3-codex/
- 2026-02 | Robotaxi service scales past 400,000 weekly rides | Mobility & Transport | Waymo | Evidence: Real-world deployment | Metric: 400,000+ weekly rides | Waymo reported substantial growth in driverless ride-hailing across major metro markets. | Possible: More people can use fully driverless transport as an actual service. | Limit: Availability remains geographically limited and depends on off-site operations. | Source: https://waymo.com/blog/2026/02/waymo-raises-usd16-billion-investment-round/
- 2026-01-13 | Vertical AI-generated video targets mobile creators | Creative Media | Google | Evidence: Product capability | Metric: Portrait outputs with upscaling | Google updated Veo 3.1 with portrait-oriented creation and improved control. | Possible: Creators can produce short-form video variants more directly. | Limit: Results may be misleading or visually inconsistent. | Source: https://blog.google/innovation-and-ai/technology/ai/veo-3-1-ingredients-to-video/
- 2025-12-30 | Agent-native APIs become a production pattern | Infrastructure | OpenAI | Evidence: Measured trend | Metric:  | OpenAI's 2025 developer review described the shift from prompting toward agent tools, reasoning controls and multimodal APIs. | Possible: Design small bounded automation pilots with evaluation, rollback and monitoring. | Limit: An industry retrospective is not a new model launch or proof of ROI. | Source: https://developers.openai.com/blog/openai-for-developers-2025
- 2025-12 | AI flood forecasting expands across countries | Climate & Weather | Google | Evidence: Real-world deployment | Metric: 2 billion people in covered risk areas | Google Research reported flood prediction coverage in 150 countries for significant riverine floods. | Possible: Emergency planners gain another early-warning signal for high-risk locations. | Limit: Coverage does not mean every person receives an accurate personalized warning. | Source: https://research.google/blog/google-research-2025-bolder-breakthroughs-bigger-impact/
- 2025-12 | AI-powered monsoon forecasts reach Indian farmers | Food & Agriculture | Google | Evidence: Real-world deployment | Metric: 38 million farmers received forecasts | Google and partners reported delivery of longer-range monsoon forecasting to farmers in India. | Possible: Farmers and advisers can consider new predictive signals when planning crops and irrigation. | Limit: Reach is not proof of better harvests for every farmer. | Source: https://research.google/blog/google-research-2025-bolder-breakthroughs-bigger-impact/
- 2025-12 | FireSat identifies small fires missed by other satellites | Climate & Weather | Google | Evidence: Research demonstration | Metric: Early small-fire detection | Google Research described a FireSat sensor spotting a relatively small roadside fire that other space-based systems missed. | Possible: Earlier detection may improve emergency response when systems scale. | Limit: A pilot detection does not prove a full planned constellation is operational. | Source: https://research.google/blog/google-research-2025-bolder-breakthroughs-bigger-impact/
- 2025-12 | Robotaxis serve millions of commercial passenger trips | Mobility & Transport | Waymo | Evidence: Real-world deployment | Metric: More than 14 million trips reported for 2025 | Waymo's annual report described over 14 million rides during 2025. | Possible: Driverless rides became ordinary paid transport in supported service areas. | Limit: Operator-reported trips are not universal autonomy across all weather and roads. | Source: https://waymo.com/blog/2026/02/waymo-raises-usd16-billion-investment-round/
- 2025-10-06 | Codex broadens engineering-team integrations | Coding & Agents | OpenAI | Evidence: Product capability | Metric:  | Codex general availability included SDK and team-oriented workflow options. | Possible: Connect code review and engineering workflows with auditable checkpoints. | Limit: Integrations and platform entitlements evolve; verify current documentation. | Source: https://openai.com/index/codex-now-generally-available/
- 2025-09-30 | AI video and dialogue generation become synchronized | Creative Media | OpenAI | Evidence: Research demonstration | Metric: Sora 2 synchronized sound | OpenAI launched Sora 2 with stronger motion control and synchronized audio. | Possible: The release showed growing capacity for rapid audiovisual prototyping. | Limit: Standalone Sora later closed in April 2026; this is a historical milestone, not current availability. | Source: https://openai.com/index/sora-2/
- 2025-09-29 | Claude Sonnet 4.5 expands coding and computer use | Coding & Agents | Anthropic | Evidence: Product capability | Metric:  | Anthropic reported improvements in coding, agent work and computer-use capability. | Possible: Pilot supervised coding and tool-driven tasks, checking output and permissions. | Limit: Vendor-reported performance is not a universal independent ranking. | Source: https://www.anthropic.com/news/claude-sonnet-4-5
- 2025-09-29 | Coding agents gain rollback and deeper tooling | Software & Work | Anthropic | Evidence: Product capability | Metric: Checkpoints and VS Code integration | Anthropic shipped coding-agent checkpoints and longer-context tooling. | Possible: Engineers can delegate bounded changes with review and rollback. | Limit: Provider performance claims require task-specific evaluation. | Source: https://www.anthropic.com/news/claude-sonnet-4-5
- 2025-08-28 | Real-time voice agents gain tools and calling | Accessibility & Language | OpenAI | Evidence: Product capability | Metric: SIP calling and external tools | OpenAI made Realtime API generally available with richer speech, phone and tool capabilities. | Possible: Businesses can prototype supervised voice workflows with authorized tools. | Limit: Human escalation, disclosure and account permissions are essential. | Source: https://openai.com/index/introducing-gpt-realtime/
- 2025-08-26 | Live language translation reaches phone conversations | Accessibility & Language | Google | Evidence: Product capability | Metric: Live spoken translation | Google Translate added live multilingual audio conversations. | Possible: Travelers and multilingual groups can reduce communication friction. | Limit: Translation nuance and reliability vary by language. | Source: https://blog.google/products-and-platforms/products/translate/language-learning-live-translate/
- 2025-08-07 | GPT-5 introduces built-in reasoning routing | Models | OpenAI | Evidence: Product capability | Metric:  | GPT-5 combined faster responses and longer deliberation under one model family. | Possible: Compare fast and deliberate reasoning on representative professional tasks. | Limit: This is a historical release, not a claim that GPT-5 remains the latest model. | Source: https://openai.com/index/introducing-gpt-5/
- 2025-08-05 | Text prompts create navigable interactive virtual worlds | Creative Media | Google DeepMind | Evidence: Research demonstration | Metric: 720p at 24 fps for minutes | Genie 3 demonstrated short-lived interactive worlds navigable in real time. | Possible: Research teams can prototype simulations and agent training environments. | Limit: Generated environments do not guarantee real-world physical fidelity. | Source: https://deepmind.google/blog/genie-3-a-new-frontier-for-world-models/
- 2025-07-21 | AI solutions reach gold-medal math standard | Mathematics & Algorithms | Google DeepMind | Evidence: Independent evidence | Metric: 35 of 42 IMO points | Gemini Deep Think solved five International Mathematical Olympiad problems under official grading. | Possible: Advanced mathematical reasoning has crossed a notable competition milestone. | Limit: Competition performance does not prove general mathematical research reliability. | Source: https://deepmind.google/blog/advanced-version-of-gemini-with-deep-think-officially-achieves-gold-medal-standard-at-the-international-mathematical-olympiad/
- 2025-07-17 | ChatGPT agent combines research and browser actions | Agents | OpenAI | Evidence: Product capability | Metric:  | A combined agent could research, navigate websites and work across tools during a supervised task. | Possible: Delegate multi-step research and drafting with permission boundaries and checkpoints. | Limit: Product controls, plan access and reliability change over time. | Source: https://openai.com/index/introducing-chatgpt-agent/
- 2025-07-10 | AI coding makes some expert developers slower | Software & Work | METR | Evidence: Independent evidence | Metric: 19% longer completion time | METR randomized 246 tasks across 16 experienced developers and found early-2025 AI tools increased task time. | Possible: Organizations should measure real workflow speed rather than believe every benchmark. | Limit: A small sample and early tool versions limit generalization. | Source: https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/
- 2025-07-10 | Still photos turn into short videos with audio | Creative Media | Google | Evidence: Product capability | Metric: Eight-second photo-to-video clips | Google expanded Veo-powered creation to animate images with sound. | Possible: Creators can animate rights-cleared photographs and mockups. | Limit: Synthetic footage must not be represented as recorded events. | Source: https://blog.google/products-and-platforms/products/gemini/photo-to-video/
- 2025-07-01 | Operational AI weather ensembles arrive | Climate & Weather | ECMWF | Evidence: Real-world deployment | Metric: 51 ensemble forecasts | ECMWF deployed an AI-driven probabilistic forecast ensemble in parallel with conventional methods. | Possible: Meteorologists can compare model uncertainty and multiple potential outcomes for decision making. | Limit: Lower resolution and variable-dependent performance require conventional cross-checks. | Source: https://www.ecmwf.int/en/about/media-centre/news/2025/ecmwfs-ensemble-ai-forecasts-become-operational
- 2025-06-25 | AlphaGenome begins decoding variant consequences | Biology & Medicine | Google DeepMind | Evidence: Research demonstration | Metric: Unified DNA variant-effect prediction | DeepMind released AlphaGenome to predict multiple molecular properties of genetic variants. | Possible: Researchers can better prioritize questions about gene regulation. | Limit: These are modeled effects, not clinical diagnoses. | Source: https://deepmind.google/blog/alphagenome-ai-for-better-understanding-the-genome/
- 2025-06-25 | AlphaGenome predicts effects of DNA variants | Science | Google DeepMind | Evidence: Research demonstration | Metric:  | AlphaGenome introduced predictions about regulatory effects of DNA sequence variants. | Possible: Prioritize experimental questions in genomics, subject to specialist assessment. | Limit: Predictions are not clinical diagnoses or experimentally confirmed biological effects. | Source: https://deepmind.google/blog/alphagenome-ai-for-better-understanding-the-genome/
- 2025-05-16 | Cloud coding agents work on repository tasks | Coding & Agents | OpenAI | Evidence: Product capability | Metric:  | Codex research preview could write code, investigate bugs and propose pull requests in isolated environments. | Possible: Delegate bounded test generation and code changes while reviewing diffs and CI outcomes. | Limit: Original preview capabilities and availability differ from today's Codex versions. | Source: https://openai.com/index/introducing-codex/
- 2025-05-14 | AlphaEvolve searches automatically for better algorithms | Mathematics & Algorithms | Google DeepMind | Evidence: Research demonstration | Metric: Automatic propose-and-evaluate algorithm search | DeepMind launched AlphaEvolve for discovering computational algorithms and mathematics improvements. | Possible: Teams can test candidate optimization strategies where success has a measurable objective. | Limit: Only reliably checkable problems suit this approach well. | Source: https://deepmind.google/blog/alphaevolve-a-gemini-powered-coding-agent-for-designing-advanced-algorithms/
- 2025-05 | Aurora predicts varied Earth-system phenomena | Climate & Weather | Microsoft Research | Evidence: Independent evidence | Metric: More than one million hours of geophysical data | Microsoft researchers published Aurora, an Earth-system foundation model tested on several environmental tasks. | Possible: Scientists can reuse a pretrained model across weather and environmental studies. | Limit: Research comparisons do not replace official forecasts. | Source: https://www.microsoft.com/en-us/research/publication/aurora-a-foundation-model-for-the-earth-system/
- 2025-04-10 | Electricity demand projections reveal AI capacity constraints | Infrastructure & Energy | International Energy Agency | Evidence: Measured trend | Metric: 945 TWh global data-center demand by 2030 | IEA projects more than doubling demand with AI a major driver. | Possible: Utilities need scenario-based capacity planning and permitting. | Limit: A projection is not a realized 2030 outcome. | Source: https://www.iea.org/reports/energy-and-ai/executive-summary
- 2025-04-10 | IEA measures the energy footprint behind the AI boom | Infrastructure & Energy | International Energy Agency | Evidence: Independent evidence | Metric: 415 TWh global data-center electricity in 2024 | IEA reports data-center electricity at roughly 1.5% of world consumption. | Possible: Grid planners can see the actual baseline for digital infrastructure demand. | Limit: Not all electricity in these data centers is used by AI. | Source: https://www.iea.org/reports/energy-and-ai/executive-summary
- 2025-03-25 | AI image generation becomes iterative and text-aware | Creative Media | OpenAI | Evidence: Product capability | Metric: Multimodal image editing | OpenAI introduced integrated image generation with improved text placement and visual editing. | Possible: Creative teams can rapidly prototype diagrams and marketing concepts. | Limit: Generated images can contain inaccurate details and intellectual-property risks. | Source: https://openai.com/index/introducing-4o-image-generation/
- 2025-03-12 | Gemini Robotics links multimodal AI to physical action | Robotics | Google DeepMind | Evidence: Research demonstration | Metric:  | Gemini Robotics and Robotics-ER linked visual and language reasoning with robot control in demonstrations. | Possible: Study emerging robot-assistance tasks in handling, sorting and manipulation. | Limit: Demonstration does not prove deployment safety or availability in ordinary workplaces. | Source: https://deepmind.google/blog/gemini-robotics-brings-ai-into-the-physical-world/
- 2025-02-25 | AI weather forecasts become operational | Climate & Weather | ECMWF | Evidence: Real-world deployment | Metric: Up to 1,000× lower energy per forecast | ECMWF launched the Artificial Intelligence Forecasting System alongside conventional forecasts and reported stronger performance on several metrics. | Possible: Weather agencies can run complementary global forecasts at significantly lower computational cost. | Limit: AIFS complements physics models; resolution and regional skill differ. | Source: https://www.ecmwf.int/en/about/media-centre/news/2025/ecmwfs-ai-forecasts-become-operational
- 2025-02-19 | Evo 2 learns from genetic code across domains of life | Biology & Medicine | Arc Institute | Evidence: Research demonstration | Metric: 9.3 trillion DNA tokens | Arc Institute launched a large open DNA language model for biological sequence prediction and generation. | Possible: Scientists can generate and rank sequence hypotheses computationally. | Limit: Biosafety and wet-lab confirmation remain crucial. | Source: https://arcinstitute.org/news/evo2
- 2025-02-10 | Usage evidence shows AI penetrating many occupations | Economy & Labor | Anthropic | Evidence: Measured trend | Metric: 36% of jobs used AI on at least a quarter of tasks | Anthropic analyzed anonymized Claude usage across occupational tasks. | Possible: Businesses can focus on task redesign rather than declaring whole occupations obsolete. | Limit: Single-platform use is not an unbiased cross-economy measure. | Source: https://www.anthropic.com/research/the-anthropic-economic-index
- 2025-02-02 | Deep research assembles multi-source reports | Research | OpenAI | Evidence: Product capability | Metric:  | Deep research introduced an agentic workflow that searches and synthesizes external material into sourced reports. | Possible: Ask for comparative research with traceable citations, then verify the important claims. | Limit: Research outputs can miss sources and include citation errors. | Source: https://openai.com/index/introducing-deep-research/
- 2025-02 | Virtual-cell datasets begin supporting predictive biology | Biology & Medicine | Arc Institute | Evidence: Research demonstration | Metric: More than 600 million cells in expanding atlas | Arc's Virtual Cell Atlas combines observations and perturbation records for training biological models. | Possible: Researchers can study how cellular systems react to interventions. | Limit: Dataset quantity does not establish causal validity or experimental generalization. | Source: https://arcinstitute.org/tools/virtualcellatlas
- 2025-01-23 | Browser agents begin performing web tasks | Agents | OpenAI | Evidence: Research demonstration | Metric:  | Operator showed an AI agent using a browser to click, type and navigate websites. | Possible: Experiment with supervised browser workflows; require explicit confirmations for purchases and sensitive actions. | Limit: Standalone Operator later moved into ChatGPT agent; historical launch, not a currently separate product. | Source: https://openai.com/index/introducing-operator/
- 2025 | Academia remains central to scientific AI infrastructure | Science & Discovery | Stanford HAI | Evidence: Measured trend | Metric: Academic and public-sector labs produce many science models | Stanford finds scientific AI model development differs from commercial frontier-model concentration. | Possible: University and public research partnerships remain important. | Limit: Patterns vary widely between different scientific disciplines. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/science
- 2025 | AI becomes a mass-population technology faster than the internet | Economy & Labor | Stanford HAI | Evidence: Measured trend | Metric: 53% adoption in three years | Stanford reports unusually rapid public adoption of generative AI across studied countries. | Possible: Governments and employers must adapt faster than earlier technology transitions. | Limit: Country variation is wide; adoption is not proof of productivity. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/economy
- 2025 | AI begins scoring strongly on select professional evaluations | Models & Benchmarks | Stanford HAI | Evidence: Measured trend | Metric: Roughly 60%–90% on selected tax, legal and finance evaluations | Stanford surveys results showing models can answer some domain exams impressively. | Possible: AI can assist with first drafts of regulated work under qualified review. | Limit: Benchmark outcomes cannot substitute for professional liability and actual case quality. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/technical-performance
- 2025 | AI computing power demands concentrate in infrastructure hubs | Infrastructure & Energy | Stanford HAI | Evidence: Measured trend | Metric: US hosts more than 5,400 data centers | Stanford describes concentrated data-center geography and key chip supply bottlenecks. | Possible: Electricity, cooling, land and hardware capacity now constrain expansion. | Limit: The number includes non-AI data centers. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report
- 2025 | AI education becomes a national curriculum requirement in some countries | Education & Learning | Stanford HAI | Evidence: Measured trend | Metric: China and UAE 2025–26 mandates | Stanford reports national-level AI education requirements entering school systems. | Possible: AI literacy is becoming a public education objective. | Limit: Mandates do not guarantee access or learning outcomes. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/education
- 2025 | AI investment more than doubles | Economy & Labor | Stanford HAI | Evidence: Measured trend | Metric: About $581.7 billion total corporate AI investment | Stanford reports exceptional 2025 global spending, with private funding and mergers driving growth. | Possible: Capital allocation, research priorities and chip demand change at scale. | Limit: Investment is not equivalent to economic returns. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/economy
- 2025 | AI sovereignty drives government infrastructure investment | Infrastructure & Energy | Stanford HAI | Evidence: Measured trend | Metric: 44 public AI clusters in Europe and Central Asia | Stanford reports state-backed compute expansion and uneven global capacity distribution. | Possible: Access to computing becomes a national strategy consideration. | Limit: Cluster counts do not equal comparable useful compute power. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/policy-and-governance
- 2025 | AI videos begin exhibiting surprising physical patterns | Creative Media | Stanford HAI | Evidence: Measured trend | Metric: 18,000+ AI videos examined in a study | Stanford reviews work finding some video models demonstrate zero-shot physical-pattern reasoning. | Possible: Generated environments may become useful in simulation and training. | Limit: Physical plausibility is not reliable simulation of reality. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/technical-performance
- 2025 | AI-generated answers dominate many health searches | Search & Information | Stanford HAI | Evidence: Measured trend | Metric: 84%–92% of health queries showed AI summaries in studied samples | Stanford reviews studies of generated search summaries appearing before traditional results. | Possible: Health information publishers and patients now encounter machine summaries as a primary interface. | Limit: Prevalence is not accuracy or medical advice quality. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/medicine
- 2025 | AI-written scientific paper reaches workshop acceptance | Science & Discovery | Stanford HAI | Evidence: Research demonstration | Metric: One AI-generated paper accepted at an ICLR workshop | Stanford describes AI Scientist-v2 reaching a workshop peer-review milestone. | Possible: AI can automate fragments of idea generation, experimentation and paper writing. | Limit: One accepted workshop paper is not evidence of reliable autonomous science. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/science
- 2025 | Astronomy gains foundation models across giant sky surveys | Science & Discovery | Stanford HAI | Evidence: Measured trend | Metric: More than 200 million celestial objects in one model training set | Stanford describes AION-1 and associated astronomical datasets and benchmarks. | Possible: Astronomers can explore cross-survey signal retrieval and classification. | Limit: Training scale does not establish discovery quality. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/science
- 2025 | Bioinformatics agents miss most real-world analysis tasks | Science & Discovery | Stanford HAI | Evidence: Measured trend | Metric: Around 17% accuracy in BixBench | Stanford highlights major limitations in automated bioinformatics tasks. | Possible: Researchers can identify workflows where verification remains indispensable. | Limit: AI can confidently generate wrong analysis outputs. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/science
- 2025 | Chemistry models exceed average expert test scores | Science & Discovery | Stanford HAI | Evidence: Measured trend | Metric: 2,700+ ChemBench questions | Stanford highlights chemistry-question performance reaching above average expert results. | Possible: AI can accelerate scientific question triage and literature interrogation. | Limit: Strong exam performance does not imply experimental replicability. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/science
- 2025 | Chinese and US models compete near the same frontier | Models & Benchmarks | Stanford HAI | Evidence: Measured trend | Metric: Top performance gap falls to single digits | Stanford documents a narrowing capability gap between US and Chinese providers. | Possible: Organizations can assess more model alternatives and regional options. | Limit: Arena-style ratings are a particular benchmark, not universal intelligence. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/technical-performance
- 2025 | Clinical AI scribes reach hospital-scale use | Biology & Medicine | Stanford HAI | Evidence: Measured trend | Metric: Up to 83% reduction in note-writing time in cases reviewed | Stanford describes large health-system use of ambient documentation tools. | Possible: Clinicians can spend less time on routine paperwork if edits are accurately reviewed. | Limit: Reported benefits vary and note inaccuracies may harm care. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/medicine
- 2025 | Clinical proof lags behind AI device deployment | Biology & Medicine | Stanford HAI | Evidence: Measured trend | Metric: 2.4% of reviewed supporting studies had randomized trials | Stanford highlights gaps in prospective evidence supporting many medical AI authorizations. | Possible: Health systems can require stronger patient-level evaluation before adoption. | Limit: Reported fraction is conditional on the reviewed studies, not all devices. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/medicine
- 2025 | Computer-use agents make enormous benchmark gains | Models & Benchmarks | Stanford HAI | Evidence: Measured trend | Metric: About 12% to 66.3% success on OSWorld | Stanford tracks major improvements in AI agents performing cross-application computer tasks. | Possible: Computer agents can tackle more supervised office and browsing workflows. | Limit: They still fail about one in three structured benchmark attempts. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/technical-performance
- 2025 | Diagnostic agents excel on complex simulated cases | Biology & Medicine | Stanford HAI | Evidence: Measured trend | Metric: Up to 85.5% accuracy on a selected case benchmark | Stanford examines a multi-agent medical diagnostic system evaluated on published clinical vignettes. | Possible: Research can test multi-agent diagnostic support under supervision. | Limit: Unaided-physician comparisons and simulated cases do not prove superior real-world care. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/medicine
- 2025 | Digital-twin diabetes intervention shows early results | Biology & Medicine | Stanford HAI | Evidence: Independent evidence | Metric: 150-participant randomized study | Stanford describes an early intervention trial reporting improvements in blood-sugar control. | Possible: Personalized digital-twin interventions warrant larger controlled trials. | Limit: The finding cannot be extrapolated to all conditions or patients. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/medicine
- 2025 | End-to-end learned weather pipelines emerge | Climate & Weather | Stanford HAI | Evidence: Measured trend | Metric: Aardvark integrated forecasting research | Stanford reports research toward replacing complicated prediction-pipeline stages with AI. | Possible: Lower-cost forecasting architectures become viable research directions. | Limit: Operational meteorology still requires verification and contingency planning. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/science
- 2025 | English-language AI quality masks global dialect failures | Accessibility & Language | Stanford HAI | Evidence: Measured trend | Metric: Large performance gaps for regional dialects | Stanford documents weaker results in some non-English and regional language contexts. | Possible: Organizations should test locally relevant language varieties directly. | Limit: Strong English benchmarks do not guarantee multilingual reliability. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/responsible-ai
- 2025 | Estimated consumer benefits of generative AI rise | Economy & Labor | Stanford HAI | Evidence: Measured trend | Metric: $172 billion annual US consumer surplus estimated by early 2026 | Stanford describes rapid growth in estimated value consumers derive from low-cost AI tools. | Possible: People are gaining utility outside paid enterprise software. | Limit: Consumer surplus is an estimate, not direct revenue or wage gains. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/economy
- 2025 | Familiar AI benchmarks begin to saturate | Models & Benchmarks | Stanford HAI | Evidence: Measured trend | Metric: As many as 42% invalid test questions in one reviewed benchmark | Stanford documents benchmark contamination and quality issues. | Possible: Businesses need independent tests on actual tasks rather than buying by a leaderboard rank. | Limit: A score jump does not necessarily mean better deployment performance. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/technical-performance
- 2025 | Frontier training compute grows at an extraordinary pace | Infrastructure & Energy | Stanford HAI | Evidence: Measured trend | Metric: Historical estimate around 3.3× annually | Stanford reports rapid growth in training compute since 2022. | Possible: Infrastructure concentration may shape who can compete at the frontier. | Limit: Historical trends are not guaranteed to persist. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report
- 2025 | Hallucination remains common under certain tests | Safety, Trust & Policy | Stanford HAI | Evidence: Measured trend | Metric: 22%–94% across models on one test | Stanford documents huge variation in model behavior when handling false user-supplied beliefs. | Possible: High-stakes AI outputs need independent verification. | Limit: One adversarial evaluation cannot summarize general factuality. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/responsible-ai
- 2025 | Household robots remain much less capable than demos suggest | Robotics | Stanford HAI | Evidence: Measured trend | Metric: About 12% real household task success | Stanford contrasts strong robotics simulation performance with poor results in real homes. | Possible: Practical physical automation should start in predictable environments. | Limit: Viral robotics clips do not establish general household autonomy. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/technical-performance
- 2025 | Model-maker transparency declines in an index | Safety, Trust & Policy | Stanford HAI | Evidence: Measured trend | Metric: Average score declines to 40 | Stanford reports worsening disclosure of model data, training and post-deployment effects. | Possible: External assurance becomes harder as systems scale. | Limit: Index scores depend on measurement criteria. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/responsible-ai
- 2025 | Most high-school and college students use AI for schoolwork | Education & Learning | Stanford HAI | Evidence: Measured trend | Metric: More than 80% reported use | Stanford describes widespread AI use for homework, writing and research. | Possible: Education systems need source literacy, assessment redesign and AI instruction. | Limit: Use does not prove learning gains. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/education
- 2025 | Much clinical-AI testing still relies on simulated patients | Biology & Medicine | Stanford HAI | Evidence: Measured trend | Metric: Nearly half of clinical AI studies use simulated scenarios | Stanford identifies a gap between published diagnostic demonstrations and patient-level trials. | Possible: Clinical developers need prospective evaluations that reflect real clinics. | Limit: Excellent simulation accuracy is not a patient outcome. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/medicine
- 2025 | National AI strategies spread across emerging economies | Safety, Trust & Policy | Stanford HAI | Evidence: Measured trend | Metric: New national planning across more regions | Stanford documents government strategy adoption outside early AI powers. | Possible: AI affects education, national infrastructure and economic policy worldwide. | Limit: Policy plans differ from executed projects. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/policy-and-governance
- 2025 | Olympiad reasoning coexists with basic perceptual failures | Models & Benchmarks | Stanford HAI | Evidence: Measured trend | Metric: Around 50.6% top-model analog-clock reading accuracy | Stanford contrasts extraordinary competition math with brittle performance on ordinary perception. | Possible: AI evaluations should include mundane tests where mistakes matter. | Limit: The abilities of AI remain sharply uneven across task types. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/technical-performance
- 2025 | Only a small minority of teachers find AI policies clear | Education & Learning | Stanford HAI | Evidence: Measured trend | Metric: About 6% report clear school AI rules | Stanford finds substantial ambiguity around what students and teachers can do with AI. | Possible: Clear policies and teacher training are urgent school operations needs. | Limit: Survey findings vary by district and sample. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/education
- 2025 | Organizational use spreads while fully autonomous agents lag | Economy & Labor | Stanford HAI | Evidence: Measured trend | Metric: 88% of surveyed organizations use AI | Stanford reports AI use in at least one function becoming commonplace while agent adoption remains single-digit in many functions. | Possible: Organizations can redesign routine tasks but should measure outcomes separately. | Limit: Use does not imply profitability or fully automated processes. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/economy
- 2025 | Public confidence trails expert enthusiasm about AI | Society & Culture | Stanford HAI | Evidence: Measured trend | Metric: 73% of experts vs 23% public expect positive work impacts | Stanford describes a major difference in sentiment about AI and employment. | Possible: Institutions will need transparent evidence about what AI actually changes. | Limit: Opinions are not forecasts and depend on survey wording. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report
- 2025 | Reported AI harms grow quickly | Safety, Trust & Policy | Stanford HAI | Evidence: Measured trend | Metric: 362 recorded incidents compared with 233 in 2024 | Stanford summarizes reported AI failures and misuse incidents. | Possible: A growing incident corpus can support more systematic response and governance. | Limit: Counts are affected by media attention and reporting rates. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/responsible-ai
- 2025 | Research-replication agents remain unreliable | Science & Discovery | Stanford HAI | Evidence: Measured trend | Metric: Below 20% on paper-scale replication tests | Stanford documents low success rates on reproducing research workflows. | Possible: Scientists should demand repeatable experiments, not persuasive AI-written manuscripts. | Limit: A good-looking paper may hide failed research steps. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/science
- 2025 | Responsible-AI governance becomes an organizational function | Safety, Trust & Policy | Stanford HAI | Evidence: Measured trend | Metric: 17% more dedicated AI governance roles | Stanford reports growing adoption of policy frameworks and internal oversight positions. | Possible: AI governance is moving into operations and compliance. | Limit: Having a role does not imply effective enforcement. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/responsible-ai
- 2025 | Safety defenses weaken under adversarial prompts | Safety, Trust & Policy | Stanford HAI | Evidence: Measured trend | Metric: Safety benchmark scores degrade under attack | Stanford reviews jailbreak results and safety limitations. | Possible: Agentic systems need defense-in-depth and access controls outside the model. | Limit: Performance depends on system architecture and attack technique. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/responsible-ai
- 2025 | Scientific agents trail expert PhDs on complex research tasks | Science & Discovery | Stanford HAI | Evidence: Measured trend | Metric: 38.8% versus 83.5% expert baseline on PaperArena | Stanford reports a large gap between frontier systems and domain experts. | Possible: AI can assist experts without replacing their verification. | Limit: One benchmark cannot describe every area of science. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/science
- 2025 | Scientific AI publications surge across disciplines | Science & Discovery | Stanford HAI | Evidence: Measured trend | Metric: Around 80,150 natural-science AI papers in 2025 | Stanford reports a 26% year-over-year increase in scientific AI publications. | Possible: AI-assisted research is now a broad scientific activity rather than an isolated niche. | Limit: Paper volume does not equal validated discoveries. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/science
- 2025 | Small specialized biology models beat giant systems on some tasks | Biology & Medicine | Stanford HAI | Evidence: Measured trend | Metric: 111-million parameter model exceeds some larger baselines | Stanford notes domain-specific models outperforming much larger systems on selected protein tests. | Possible: Small specialist models may be more efficient than a generic frontier model. | Limit: Ranking depends on task, data and experimental setup. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/medicine
- 2025 | University computer-science enrollment shifts | Education & Learning | Stanford HAI | Evidence: Measured trend | Metric: 11% drop at US four-year institutions | Stanford reports lower US CS enrollment alongside growth in specialist AI graduate programs. | Possible: Curricula and career planning should distinguish engineering fundamentals from AI specializations. | Limit: Enrollment change cannot be attributed solely to AI. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/education
- 2025 | US and China competition spans research, patents and models | Infrastructure & Energy | Stanford HAI | Evidence: Measured trend | Metric: China leads in some patent and publication counts | Stanford notes growing global scientific competition despite concentrated frontier training. | Possible: Companies should consider multiple ecosystems for research and innovation. | Limit: Quantity is not equivalent to quality or adoption. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report
- 2025 | US authorizes hundreds of AI-enabled medical devices | Biology & Medicine | Stanford HAI | Evidence: Independent evidence | Metric: 258 FDA device authorizations | Stanford reports a growing regulated medical-device market for AI functionality. | Possible: More specialized AI products can be evaluated within existing regulatory pathways. | Limit: Authorization is not proof of improved patient outcomes in every setting. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/medicine
- 2025 | Virtual-cell models begin forecasting interventions | Biology & Medicine | Stanford HAI | Evidence: Measured trend | Metric: Evo 2, STATE and AlphaGenome among examples | Stanford describes biological models that try to predict molecular and cellular responses. | Possible: Scientists can narrow experimental search before laboratory testing. | Limit: Predictions are not verified treatment effects. | Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/medicine

## Methodology
- Each entry describes a distinct capability, reported outcome, study finding or historical release; none is a claim that AI Nextia independently reproduced the result.
- A product release means a tool exists or was demonstrated, not that it is universally available or a proven productivity improvement.
- Operational deployments and independently measured studies are labeled separately from laboratory demonstrations and vendor claims.
- Year-wide 2025 statistics are grouped under 2025 Annual Findings without fabricating an exact announcement month.
- Dates of individual releases are taken from original sources where possible; historical records are not proof of current product availability.
- Source-linked adoption figures and modeled forecasts retain their original geographic, sample and methodological limitations.
- The 2027+ watchlist is not a list of achievements or promised release dates.
- The catalog is curated and growing; it is not an exhaustive index of every AI model, paper, company or social-media demonstration.

## Practical guides
- [Summarize meetings into tracked decisions](https://ainextia.com/what-ai-can-do/summarize-meetings/): Turn a consented meeting transcript into a concise summary, decisions, owners, deadlines, and unanswered questions.
- [Find recurring themes in customer feedback](https://ainextia.com/what-ai-can-do/customer-feedback/): Group consented reviews and support messages into themes and possible operational issues.
- [Extract fields from invoices and forms](https://ainextia.com/what-ai-can-do/document-extraction/): Use document recognition and language models to extract structured fields for human verification.
- [Ask questions of approved internal documentation](https://ainextia.com/what-ai-can-do/knowledge-assistant/): Build a source-grounded assistant over authorized documentation, with linked citations and explicit unknowns.
- [Draft structured client proposals](https://ainextia.com/what-ai-can-do/draft-proposals/): Use approved scope, pricing, client context, and a standard template to propose next steps.
- [Prioritize inbound support requests](https://ainextia.com/what-ai-can-do/support-triage/): Classify requests and route them to the right team with clear human overrides.
- [Generate focused regression tests from changes](https://ainextia.com/what-ai-can-do/code-tests/): Use AI coding assistants to propose tests based on a precise code diff and known behavior.
- [Turn code changes into human-readable release notes](https://ainextia.com/what-ai-can-do/release-notes/): Summarize merged pull requests and release commits into verified notes.
- [Build a cited literature map](https://ainextia.com/what-ai-can-do/research-literature/): Search papers systematically and extract supported claims, methods, and disagreements.
- [Create controlled variations of a visual concept](https://ainextia.com/what-ai-can-do/image-iterations/): Iterate on licensed imagery and approved brand assets while keeping core product details consistent.
- [Turn data exports into a reviewed weekly narrative](https://ainextia.com/what-ai-can-do/ai-reporting/): Combine structured, dated exports with a controlled reporting template.
- [Pilot a human-approved AI workflow agent](https://ainextia.com/what-ai-can-do/agent-workflow/): Use a small sandboxed agent to interpret inputs and draft actions that require approval.
- [Reconcile invoices and payments](https://ainextia.com/what-ai-can-do/invoice-reconciliation/): Compare invoice, purchase-order and payment records; flag duplicates, missing payments and mismatched amounts for finance review.
- [Draft customer email responses from policies](https://ainextia.com/what-ai-can-do/email-response-drafts/): Use approved policy documents to prepare draft customer answers with citations for staff approval.
- [Create a marketing campaign brief](https://ainextia.com/what-ai-can-do/marketing-campaign-brief/): Draft audience hypotheses, messaging and campaign concepts based on approved brand assets and research.
- [Clean messy spreadsheets and flag duplicates](https://ainextia.com/what-ai-can-do/spreadsheet-data-cleanup/): Standardize spreadsheets, normalize fields and flag anomalies while preserving original records.
- [Research competitors with cited evidence](https://ainextia.com/what-ai-can-do/research-competitors/): Gather product positioning, pricing, public claims and customer evidence from original websites, marking uncertain facts.
- [Find key scenes inside long videos](https://ainextia.com/what-ai-can-do/video-scene-index/): Use multimodal AI to identify scenes, themes and time codes in videos you have rights to analyze.
- [Create short animated concept loops](https://ainextia.com/what-ai-can-do/animated-product-concepts/): Generate short looping GIF concepts from approved images and review for brand consistency.
- [Translate and quality-check multilingual content](https://ainextia.com/what-ai-can-do/translation-proofreading/): Produce translations of approved copy while checking terminology, instructions and cultural context.
- [Explore candidate genomic variant effects](https://ainextia.com/what-ai-can-do/research-dna-variants/): Review predicted variant effects and relevant literature in authorized scientific research.
- [Evaluate autonomous agents before production](https://ainextia.com/what-ai-can-do/evaluate-agent-workflow/): Create representative agent tasks, score outcomes, track cost and inspect erroneous tool actions.
- [Produce an evidence-backed research report](https://ainextia.com/what-ai-can-do/build-cited-report/): Synthesize a literature set or source collection with citations and uncertainty rather than unsupported claims.
- [Prototype an accessible web interface with AI coding](https://ainextia.com/what-ai-can-do/prototype-web-interface/): Use AI coding tools for a scoped accessible interface prototype with tests and manual review.
- [Prototype a supervised voice support assistant](https://ainextia.com/what-ai-can-do/voice-support-draft/): Connect speech input and output to an approved FAQ workflow, with explicit consent and handoff.
- [Prioritize scientific papers for expert review](https://ainextia.com/what-ai-can-do/scientific-literature-triage/): Extract methods, sample characteristics and limitations from a set of original papers.
- [Assist authorized security fix verification](https://ainextia.com/what-ai-can-do/security-fix-validation/): Use an approved security assistant to organize defensive findings and validate that fixes address known issues.
- [Explain weather model outputs with uncertainty](https://ainextia.com/what-ai-can-do/forecast-explanation/): Compare weather-model outputs, confidence and regional differences for an expert-facing briefing.

## Future watchlist — not announced discoveries
- 2027+ | AI agents completing reliable multi-app projects | Evidence needed: Production evaluations and independently documented success/failure rates. | Not a forecast that reliable full autonomy is achieved in 2027.
- 2027+ | More accessible physical robotics | Evidence needed: Independent field studies and repeatable robotics benchmarks. | Do not infer commercial readiness from highlight reels.
- 2027+ | Verifiable provenance for generated answers | Evidence needed: Auditable protocols, deployments and published third-party evaluations. | Research prototypes may not become widespread products.
- 2027+ | AI-supported scientific experiments | Evidence needed: Peer-reviewed replications and controls. | AI-generated hypotheses are not discoveries until confirmed.
- 2027+ | AI media with longer temporal consistency | Evidence needed: Benchmarked samples and published access conditions. | Visual demos may omit failures or editing steps.
- 2027+ | Outcome-based AI automation benchmarks | Evidence needed: Independent ROI reports with full cost accounting. | No guaranteed productivity increase is implied.
