Skip to main content

Reviews, news and how-to guides

AI articles and reviews — page 24

Find a review, catch up on the news or learn how to complete a task. Browse by topic or filter the library below.

Not sure where to begin?

Answer three questions and get a focused tool shortlist for your budget and priority.

Use the AI Tool Finder →

Choose a pathway

Explore by topic

Editor recommended

Recently published

Latest articles

Browse everything ↓

The weekly briefing

Make better AI decisions each week.

Get tool recommendations, product and pricing news, and a practical task to try each week.

Browse reviews, news and guides

1173 articles · Page 24 of 66

Clear filters
Paper-cut editorial illustration of one personal AI agent branching into five uses for research, calendar and email, travel, shopping, and recurring projects
GuideWork & Operations

Top 5 Meta Muse Uses: Practical Tasks Worth Testing First

The best first uses are valuable enough to save time, bounded enough to review, and reversible when the agent gets something wrong.

Start Muse with research, inbox and calendar triage, travel planning, comparison shopping, and recurring projects—then expand authority only from measured results.

Read guide

Paper-cut editorial illustration of an orbital AI chip satellite passing through radiation, cooling, power, laser-link, and economics checkpoints above Earth
GuideWork & Operations

Google Is Sending a TPU to Orbit—What Project Suncatcher Must Prove

The first orbital hardware test can answer survival and cooling questions, but it is a long way from proving that space-based AI compute is economical or scalable.

Google is preparing an orbital TPU test for Project Suncatcher, its research effort to explore solar-powered AI compute in space.

Read guide

Paper-cut editorial illustration of an AI security hypothesis entering an isolated test chamber and becoming a verified exploit before human review
Guide

Google PageBreak Makes AI Prove the Exploit Before Raising the Alarm

PageBreak's important idea is not another vulnerability hypothesis—it is a deterministic validator that must reproduce the exploit before a product team sees it.

Google says its internal PageBreak agent found more than 500 XSS vulnerabilities while keeping speculative AI findings away from product teams.

Read guide

Paper-cut editorial illustration of a warehouse robot splitting fast local safety controls from a larger remote AI model through a measured edge-compute path with latency and failover gates
GuideWork & Operations

Microsoft Moves Robot AI Offboard—Latency Becomes a Safety Boundary

Remote GPUs can unlock larger models and longer battery life, but a network path must never become an unexamined control dependency.

Microsoft Research reported that moving physical-AI inference from onboard robot GPUs to nearby edge or cloud compute improved task success in representative mobile-manipulation workloads, supported larger models, and could extend battery life. It also added offloaded inference to the open-source Physical AI Toolchain. The opportunity is more capable robots with lighter onboard compute; the engineering constraint is that latency, jitter, bandwidth, loss, and remote availability become part of the robot's safety case.

Read guide

Paper-cut editorial illustration of a target molecule branching backward into candidate reaction routes that pass through literature, feasibility, hazard, and expert laboratory approval gates
GuideWork & Operations

RetroChimera Maps How to Make Molecules—Chemists Still Hold the Gate

The open model can widen retrosynthesis search, but its own maintainers warn that predictions can hallucinate and require expert verification.

Microsoft Research, GSK, Novartis, and collaborators introduced RetroChimera, an open retrosynthesis model that works backward from a desired small molecule to propose reactions that might make it. Microsoft reports stronger benchmark and chemist-preference results than comparison systems, plus adaptation to proprietary chemistry data. The opportunity is faster candidate planning—not autonomous drug creation. Suggested reactions can be wrong, out of distribution, unsafe, impractical, or economically useless.

Read guide

Paper-cut editorial illustration of buyer questions flowing through multiple answer engines into citations, crawler visits, verified analytics, and a guarded content action loop
ReviewContent & Search

Profound Review 2026: AI Search Visibility, Agents & Pricing

A research-based assessment of Profound's capabilities, economics, evidence, and operational fit.

Profound is a serious shortlist for established brands building a measurable answer-engine optimization program. It connects prompt demand, brand visibility, citations, crawler behavior, AI-referred traffic, and content workflows. Its biggest limitation is attribution: observed mentions, citations, and visits do not by themselves prove incremental revenue or that a Profound recommendation caused a change.

Read guide

Paper-cut editorial illustration of voice and chat conversations entering a governed contact center where AI agents, assisted humans, quality review, and customer escalation remain connected
ReviewWork & Operations

Cresta Review 2026: Contact Center AI, Agents & Pricing

A research-based assessment of Cresta's capabilities, economics, evidence, and operational fit.

Cresta is worth evaluating for large contact centers that want AI agents, real-time human-agent guidance, and conversation intelligence on one enterprise platform. The opportunity is a shared learning loop across automated and human conversations. The risk is optimizing a vendor score or containment rate while customer outcomes, consent, fairness, or escalation quality deteriorate.

Read guide

Paper-cut editorial illustration of many AI research scouts narrowing a vast DNA sequence field into one candidate that passes through expert review and a physical laboratory validation gate
GuideWork & Operations

Claude Found an Enzyme System—The Real Story Is the Research Loop

The result suggests AI can widen hypothesis search, but the enzyme's function is still unknown and human laboratory validation remains decisive.

Anthropic reported that roughly 950 Claude agents used 210 million tokens over 21 hours to search genomic data, identify a previously uncharacterized array-associated reverse transcriptase system, and narrow a large candidate set for human review. Scientists then performed laboratory work showing that its repeat array is expressed as short RNAs. The system's biological function is still unknown, the work is an early preprint from Anthropic's own lab, and resemblance to CRISPR does not make it a gene-editing tool.

Read guide

Paper-cut editorial illustration of a conversational assistant approaching distinct productivity, creative, and consumer apps through separate permission gates with preview, approval, audit, and revoke controls
GuideBuild, Design & Govern

Gemini Adds Connected Apps—Convenience Now Depends on Permission Design

Google's new integrations can reduce tab switching, but every connection expands the data and action boundary buyers must understand.

Google began rolling out a new wave of Gemini Connected Apps on September 23, 2026, spanning productivity, creative, and consumer services. The list includes Airtable, Linear, monday.com, PandaDoc, Wispr AI, Zoho, Adobe, Picsart, Squarespace, Webflow, Apartments.com, Experian, Peloton, and SeatGeek. Users connect apps in Gemini settings and can invoke them with an @ mention or natural-language request. Availability and exact actions can differ by account, region, app, and rollout stage.

Read guide

Paper-cut editorial illustration of an expert's books, talks, and lessons entering a labeled digital knowledge twin while citations, consent, uncertainty, and human escalation remain visible
ReviewWork & Operations

Delphi AI Review 2026: Digital Minds, Voice & Pricing

A research-based assessment of Delphi's capabilities, economics, evidence, and operational fit.

Delphi is worth testing for creators, educators, coaches, and experts who have a substantial body of owned material and a clear reason to make it conversational at scale. Its opportunity is persistent audience access to an expert-shaped knowledge experience; its central risk is mistaken authority. A convincing voice and familiar answer style can make unsupported output feel more trustworthy than it is.

Read guide

Paper-cut editorial illustration of an AI agent rehearsing proposed actions in a transparent parallel lane beside live human work before passing staged approval gates toward limited write access
GuideWork & Operations

Run AI Agents in Shadow Mode Before You Grant Write Access

Let the agent observe and propose before it can send, change, delete, purchase, or trigger anything.

Shadow mode lets an AI agent receive the same inputs as a live workflow and produce proposed decisions or actions without executing them. Compare those proposals with what authorized humans actually did, score disagreements and missed cases, and grant narrowly scoped write access only after the agent meets predefined quality and safety thresholds across representative traffic.

Read guide

Paper-cut editorial illustration of an AI-assisted decision unfolding as a traceable chain of inputs, model and prompt versions, sources, tools, human approval, outcome, correction, and retention controls
GuideContent & Search

Build an AI Decision Log That Makes Outputs Reproducible

If an important AI-assisted decision cannot be reconstructed, it cannot be audited, improved, or defended reliably.

An AI decision log records enough context to explain and reconstruct an important AI-assisted outcome: case ID, purpose, input references, model and prompt versions, retrieved sources, tool calls, output, uncertainty, policy checks, human reviewer, final action, later outcome, corrections, and retention date. Store references or hashes instead of duplicating sensitive content whenever possible.

Read guide

Paper-cut editorial illustration of two remote colleagues meeting through layered dimensional glass with travel, accessibility, cost, and privacy checkpoints
GuideWork & Operations

Google Beam Expands 3D Video Meetings to Six Countries

HP Dimension with Google Beam is moving beyond pilots, but buyers still need to test whether a specialized 3D room produces enough measurable value to justify its footprint and cost.

Google Beam is expanding its headset-free 3D meeting platform to six countries, 18 channel partners, and bookable Industrious locations. The rollout creates a practical trial path for distributed teams, while Google's internal results still need independent validation.

Read guide

Paper-cut editorial illustration of multilingual speech waves moving through noise filters into timestamped transcript cards and a human review checkpoint
GuideVideo, Audio & Creative

Google Launches Gemini 3.5 Transcribe for Real-Time Voice Apps

Google's dedicated speech-to-text model targets voice agents, captions, and call analytics, but language count and vendor accuracy claims are not substitutes for testing real audio.

Gemini 3.5 Transcribe is a dedicated real-time speech-to-text model for more than 85 languages. It gives developers a new path for voice agents and live captions, while production adoption still depends on domain accuracy, latency, privacy, and cost per accepted transcript.

Read guide

Paper-cut editorial illustration of a frontier model moving through cost, speed, cache, safety, permission, evaluation, and rollback gates before entering a production agent
Guide

Claude Opus 5.5 Cuts Agent Costs—But Buyers Still Need Their Own Evals

Anthropic says its newest Opus model is faster, clearer, and roughly 40% cheaper on typical work, but internal and partner results are not a production acceptance test.

Anthropic launched Claude Opus 5.5 on September 22, 2026 across Claude, its API, AWS, Google Cloud, and Microsoft Azure. API pricing is $4 per million input tokens and $20 per million output tokens, with cache reads at $0.20. Anthropic says typical token-billed work costs about 40% less than Opus 5 and output is more than 30% faster. Those are useful buying signals, not proof that a specific agent is more reliable or cheaper end to end.

Read guide

Paper-cut editorial illustration of trusted study sources branching into a live conversation, audio notes, quizzes, flashcards, and short videos before an independent retention check
GuideVideo, Audio & Creative

Gemini Notebook Adds Live Study Conversations, Quizzes and Short Videos

The launch turns uploaded course material into a more interactive study environment, but source grounding does not guarantee correct teaching or durable learning.

Google announced new Gemini Notebook study features on September 15, 2026: real-time voice conversations grounded in notebook sources, mobile audio capture, interactive learning overviews, more quiz formats, performance follow-ups, and roughly 60-second Short Video Overviews in more than 80 languages. Rollout timing, age, language, account, and subscription limits differ by feature.

Read guide

Paper-cut editorial illustration of customer conversations, email, calendar, and pipeline records flowing into a permissioned company-memory graph before bounded agents act through approval gates
ReviewWork & Operations

Day AI Review 2026: Company Memory, GTM Agents & Pricing

A research-based review of Day AI's capabilities, economics, data boundaries, and operational fit.

Day AI combines permission-aware company memory with role-based agents that can prepare meetings, draft follow-ups, manage pipeline work, and run scheduled go-to-market tasks.

Read guide

Paper-cut editorial illustration of a meeting moving through consent, transcription, decisions, redaction, access, retention, and accountable action-item checkpoints
ReviewWork & Operations

Fellow AI Review 2026: Meeting Notes, Security & Pricing

A research-based review of Fellow's capabilities, economics, data boundaries, and operational fit.

Fellow records, transcribes, summarizes, and searches meetings while adding agendas, action items, consent capture, retention controls, integrations, and organization-wide governance.

Read guide