Tech News
← Home  ·  All topics

Agentic

214 GoKawiil briefs on this topic · RSS for agentic

Apple keeps AI optional on iPhone 18 Pro as Android leans into Gemini integration

Android Authority contrasts Google's approach of weaving Gemini AI into core Android functions like messaging and photography with Apple's more restrained AI rollout on the iPhone 18 Pro. The piece notes Apple has focused on building on-device processing to keep AI features local and private rather than pushing users toward cloud-connected tools.

GitHub adds 4M CPU cores, 160PB storage to handle AI-driven code surge

GitHub reported that over the past year it expanded its infrastructure by 4 million CPU cores, 160 petabytes of storage, and 50 Tb/s of networking capacity to cope with a surge in AI-generated code activity. The company said code commits grew sixfold and merged pull requests quadrupled year-over-year, with monthly commits rising from 1.4 billion in April to 7.4 billion by September, and total pushes over the past 12 months exceeding the combined totals of 2024 and 2025.

Security Report Details How AI Agents Already Act Inside Workplace Browsers

A security analysis outlines three methods by which AI agents currently operate within employees' browsers: remote debugging via the Chrome DevTools Protocol, browser extensions like Claude in Chrome, and agentic browsers, plus emerging standards such as WebMCP. Each method grants agents different levels of access to authenticated sessions, often indistinguishable from normal human browser activity.

Guidewire to acquire Hyperexponential for $1.7B cash

Insurance software firm Guidewire said Tuesday it will buy UK-based underwriting AI startup Hyperexponential for $1.7 billion in cash, with the deal potentially reaching $2 billion including a $400 million earnout tied to revenue targets. The acquisition is expected to close next year. Hyperexponential, founded in 2017 and backed by Andreessen Horowitz and Battery Ventures, has raised nearly $90 million and counts roughly 600 insurers across 40 countries as customers.

Bank of America warns customers against sharing login credentials with AI agents

Bank of America sent an email titled 'Protecting your finances in the AI World,' telling customers never to share their banking username and password with another person or any outside AI-enabled tool or agent acting on their behalf. The bank noted that customers remain responsible for any activity that results if they do share their credentials.

Legal expert Daniel Garrie outlines AI liability frameworks for CISOs at ISC2 Congress

At ISC2 Strategic Congress 2026 in Denver, cyber legal expert Daniel B. Garrie, CEO of Law and Forensics, presented frameworks to help CISOs justify AI deployment decisions and preserve evidence in case of disputes or breaches. The approaches include scoring AI deployment risk, documenting board-level risk decisions, and specifying what incident data must be preserved, building on JAMS AI Rules he helped author in 2024 for arbitrating AI disputes.

New plugin enforces spec-first workflow for AI coding agents before code is written

A Show HN post introduces a plugin that forces AI coding agents to write a formal specification—goals, cited file-and-line facts, steps, and done criteria—before any code is generated. The spec is checked by a script for valid file references, then reviewed by a separate agent with no prior chat context before implementation begins. Only after tests and build pass does the tool push a branch and open a pull request containing the spec.

Developer finds AI coding agent wrote dense, single-line components that passed all tests

A developer recounts a CI failure triggered by SonarQube flagging high cyclomatic complexity in a file written by the Codex AI coding agent. Inspection showed Codex had packed entire components onto single lines, but the logic was correct and appropriately handled edge cases; the developer simply instructed Codex to refactor the file into ten smaller files to satisfy the linter, and CI passed.

Warp discloses vendor breach exposing agent logs and telemetry for ~1% of users

Warp CEO Zach Lloyd disclosed that a third-party vendor in Warp's infrastructure was breached, exposing AI agent decision logs and user telemetry data. Lloyd estimates roughly 1% of Warp's over 800,000 users may be affected, and the company says it detected and mitigated the intrusion within hours of discovery on Friday. Warp has engaged CrowdStrike, notified law enforcement and affected users, and contacted the vendor, though it has not confirmed whether source code, prompts or credentials were exposed.

Android Authority writer lists feature requests for Google Wallet after redesign

Google Wallet recently switched from a horizontal scrolling card layout to a new vertically stacked design, and reports suggest further features like receipt scanning and agentic capabilities may be coming. An Android Authority columnist argues the new stacked layout makes it harder to distinguish similar-looking cards from the same bank, since it removed the alias text shown under each card previously.

Apollo economist says AI shows no profit payoff outside tech sector

Apollo Global chief economist Torsten Slok reports that excluding the 'Magnificent Seven' tech giants, the S&P 493 shows flat profit margins and revenue per worker, with no measurable productivity gains from AI adoption. He says mainstream corporations are increasingly treating AI token spending as a cost burden rather than a growth driver, even as VC-backed firms keep increasing their AI spend.

JetBrains launches ThinkRail, an agentic IDE built around isolated git workspaces

JetBrains introduced ThinkRail, a coding IDE built around the pi coding agent that runs tasks in isolated git-worktree workspaces, each with its own branch and directory. The tool maintains 'living specs' and reusable skills to retain project knowledge, manages pull request workflows including CI checks and merging, and automatically shelves finished or dormant workspaces. JetBrains says it plans to add full extension support so the IDE can build its own tools based on usage patterns.