Issue 14 · Nov 2–8, 2026Agent Age News magazine · Other editions · Method · HU / ENG

The impact of AI on software development and developers' work, curated weekly

Issue 2

Aug 10–16, 2026 · 41 stories · 1 of them from Hungary
Viewpoint

tool change · global

Claude Code's auto mode to become the default, with fewer human approvals

Anthropic has confirmed that Claude Code's auto mode will be switched on by default, so the agent will ask for permission less often. According to France's JDW, the change takes effect on 14 August.

Aug 10, 2026 · 36 sources

Who is affected and how

senior developer
Review which actions auto mode permits in your own repositories, and explicitly set the forbidden commands before the default takes effect.
tech lead
Decide at team level whether auto mode stays the default, and record the approval rules in the shared configuration.
junior developer
The agent can carry out steps without asking, so check your own changes before committing.
case · United States

SpaceX completes $60 billion acquisition of AI coding platform Cursor

SpaceX completed its record $60 billion acquisition of the AI coding platform Cursor on 14 August. The deal redraws the balance of power in the coding-tool market.

CTOReassess your Cursor dependency and vendor risk, since pricing and data handling may change after the ownership change.tech leadReview the team workflows that rely on Cursor, and keep an alternative at hand.startup founderWatch for changes to the terms, and do not lock yourself into a single coding tool.security · global

Ghostjacking and fake repos: prompt injection used to hijack coding agents

Two reports dated 11 August say the "Ghostjacking" attack and indirect prompt injection hidden in fake repos can both hijack coding agents. In one experiment, a reverse shell was opened on the…

senior developerDo not open an unfamiliar repo with an auto-mode agent before reviewing its contents.security · global

Malicious MCP servers use split instructions to make coding agents leak secrets

Researchers have shown that malicious MCP servers can split instructions into several parts to get coding agents to exfiltrate secrets (11 August). The attack bypasses simple filters.

senior developerEnable only vetted MCP servers and review their access permissions.incident · global

AI agents deleted production databases: the Replit and PocketOS cases

A summary dated 12 August reviews the Replit and PocketOS incidents, in which AI agents deleted production data. The question is what happens when an agent has access to live data.

CTORequire that no agent gets production write access without approval.

Hungary

surveyIVSZ: Hungarian software SMEs' AI use moves from ad hoc experiments to company practiceCTOCompare your own company's AI adoption maturity with the Hungarian SME survey data and identify the obstacles.startup founderUse the survey as a benchmark when planning your own use of AI tools.

More key stories of the week

tool change · global

Claude to watermark generated code and text, and developers push back

Anthropic has announced that Claude will watermark the code and text it generates. According to Hungarian reports, panic and resistance spread among developers once the details became known (13–16…

senior developerCheck that the watermark does not interfere with build, lint or licensing processes in code written by Claude.tech leadSet a rule for how the team handles marked, AI-generated code in the repository and in client deliveries.CTOAssess the legal and contractual impact of watermarking if you deliver Claude-built code to clients.startup founderConsider what marked code means for investor due diligence and IP representations.case · United States

Cognition, maker of Devin, in talks to raise funds at $40 billion valuation

Cognition, the developer of Devin, is reportedly in talks on more than $1 billion in funding at a $40 billion valuation. According to a Chinese report, this would put Devin's value above Cursor's.

CTOExpect capital to concentrate further in the market for autonomous coding agents, which will affect pricing and vendor choice.startup founderSubstantial funding strengthens competitors that sell development services as agents, so sharpen your own differentiation.tool change · global

Spotify launches Xirp, a single interface for Claude Code, Codex and Gemini CLI

Spotify unveiled Xirp on 10 August, which lets Claude Code, Codex and Gemini CLI run side by side in one place. Developers can thus manage several agents from a single tool.

senior developerTry running multiple agents in parallel in one interface, and check whether the agents work on the same file state.tech leadLook at how a multi-agent workflow can be standardised across the team.release · global

Google's Gemini 3.7 Flash targets coding and agents with a 50% introductory price cut

Google has released Gemini 3.7 Flash, a model focused on coding and agent tasks, with a 50% introductory price cut. It has also been available in GitHub Copilot since 13 August.

senior developerTest an agent task with Gemini 3.7 Flash alongside your current model and compare the quality.CTORecalculate agent costs, but keep in mind that the price cut is introductory.startup founderThe cheaper Flash model could lower the token costs of coding agents if the quality is sufficient.tool change · global

Microsoft merges Copilot into a single app, adds paid Autopilot agent tier

Microsoft is introducing a unified Copilot app that merges the consumer and enterprise versions of Copilot and adds a new paid Autopilot agent tier. The rollout began on 10–13 August, with the…

CTOAssess the licensing and data-handling implications before budgeting for the Autopilot tier.tech leadCheck how GitHub Copilot's place changes within the unified product family.release · global

Meta's Muse Code is cheaper but, by its own admission, weaker than Claude Code

Meta has unveiled Muse Code, its first coding agent, which competes with Claude Code and Codex. Meta itself acknowledged that its performance trails Claude Code's, although one test found it cheaper.

tech leadRun a benchmark on your own tasks that includes token costs, because a cheaper agent is not always cheaper in the end.CTOAdd Muse Code to your evaluation of agent vendors, but do not switch on price alone.senior developerTry it on routine tasks, where weaker performance matters less.case · European Union

Sweden's Lovable raises $400 million at a valuation of about $13 billion

Swedish vibe coding startup Lovable raised $400 million in a round valuing it at roughly $13 billion; the valuation has doubled in eight months. The news was reported between 10 and 12 August…

startup founderThe financial strength of vibe coding tools is growing, so to build an MVP quickly, compare platforms like Lovable with your own development costs.CTOPrepare for more AI-assembled applications appearing on the business side, which you will have to support or integrate.open source · Asia

DeepSeek challenges Claude Code with open-source Harness agent

On 12–13 August DeepSeek entered the coding-agent race publicly: it opened a WeChat account, is recruiting agent developers and released the open-source Harness. V4-Pro has also appeared on the API…

tech leadEvaluate Harness as an alternative, but review the source code and data handling before using it on a production codebase.CTOWeigh the privacy and compliance risks of using a Chinese vendor separately.senior developerCompare cost and quality against Claude Code on your own tasks.labour market · global

AI coding boom shifts software companies' focus to testing and verification

The AI coding boom is shifting software companies' focus towards testing and verification (10–16 August). Testing startup Blacksmith raised $45 million, and a Hungarian article argues that the more…

QA testerThe value of testing is rising, so build your skills towards automation and verification of AI-written code.tech leadScale testing and verification capacity in proportion to where AI speeds up code writing.open source · global

Torvalds calls AI code review the new normal as Linux 7.2 nears release

Linus Torvalds says AI-based code review is the new normal as the release of Linux 7.2 approaches. Another report says Linux's AI problem has no solution yet.

senior developerIf you contribute to the kernel or a large open-source project, expect AI-assisted review and the rules that apply to it.tech leadDefine the project's position on AI-generated contributions.labour market · global

AI is hollowing out the middle of software engineering, breaking the career ladder

Articles published on 13 August say AI is removing learning opportunities at every rung of the career ladder, while some people can perform above their level. InfoQ says this disrupts career…

junior developerDeliberately seek tasks where you can practise deeper coding skills, rather than relying only on the agent.tech leadPlan mentoring and hands-on assignments, because automation gives juniors fewer learning opportunities.senior developerTake an active role in developing junior colleagues.tool change · global

AI-generated code raises quality and maintainability risks, says Info-Tech; GitHub launches Code Quality

Info-Tech Research Group says AI-generated code can accelerate defects and technical debt without clear guardrails (11 August). GitHub Code Quality targets maintainability as AI-generated code…

tech leadSet up quality gates and maintainability metrics for AI-generated code.senior developerDo not accept AI code without tests and an explanation, because technical debt surfaces later.CTOIntroduce measurement of how much AI code is merged and with what defect rate.

In brief

researchResearchers find coding agents ignore open-source contribution guidelinessenior developerAs a maintainer, set up automated checks for incoming PRs rather than relying on written rules alone.tech leadIf you use an agent for open-source contributions, enforce the project's guidelines in the agent's configuration.open sourceOpen-source projects ban or restrict AI-generated code: OpenJDK, Rust, curlsenior developerRead a project's AI policy before contributing, because banned AI-generated code can lead to rejection or a ban from the project.tech leadCheck that the AI policies of the open-source dependencies you use are consistent with your own processes.securityGhostCommit exposes a blind spot in AI code reviewtech leadAdd a supplementary check alongside AI review before merging.survey34% of people shipping software with AI tools have no formal programming backgroundtech leadExpect code to come from people without a programming background too, and build review and security checks around it.senior developerPrepare to review AI-written code from contributors with non-traditional backgrounds.job marketSoftware engineering jobs hold firm despite AI layoffs, with no discernible evidence of lossessenior developerThe data show no mass layoffs, but the role is changing, so update how you work with agents.junior developerExpect the entry level to change rather than jobs to disappear.researchNew study finds 43% security flaw rate in AI coding toolstech leadIntroduce mandatory security review for changes written with AI.releaseCursor releases Origin, a Git host for fleets of AI agentstech leadAssess whether agent fleets need a separate Git host or whether your current platform is enough.releaseOpenAI and four rivals publish Agent Plugins, a common extension formatsenior developerIf you write plugins for agents, look at the format, since one package can run in several tools.tech leadFactor the format into tool selection, as it lowers switching costs.tool changeGitHub Models service shut down as developers switch to OpenAIsenior developerIf you build on the GitHub Models API, find your dependencies and plan the migration.securityCERT Intrinsec maps where autonomous coding agents leave traces on diskreleaseNew models arrive in GitHub Copilot: MAI-Code-1.1-Flash and Grok 4.6senior developerCompare the new models on your own codebase before replacing the default model.tech leadRecord which models are approved for the team, since the options differ in both cost and quality.tool changeWeekly agent roundup: Cursor reportedly building 'Sand', Claude Code learns to launch a simulatorQA testerLook at how an agent that launches the simulator on its own could be used to automate mobile and UI testing.senior developerFollow how the agents' runtime environment is expanding, and when it makes sense to involve them in debugging.tool changePlatforms integrate: Atlassian context engine, MongoDB, AWS Continuum, Semgrep–Replittech leadCheck that the context sources agents use do not leak sensitive data.tool changeDeveloper platform updates: Copilot web controls, Java SDK, JetBrains memory and ACPsenior developerIn JetBrains environments, try the agents that connect via ACP and local Ollama models.tech leadIf your team works in Java, look at the Copilot SDK for building your own tools.caseCodeRabbit raises $143 million at a $1.5 billion valuationtech leadAssess whether your team's pull request process needs a separate AI review layer, as the market is investing in this.CTOThe funding signals that reviewing AI code will become a separate cost line.tool changeStateless Model Context Protocol could reshape agent infrastructuresenior developerAvoid MCP server designs that depend on session state if the new approach becomes widespread.open sourceNVIDIA: open models and local autonomous software development with FactoryCTOFor sensitive code, consider agents running locally on open models instead of cloud-based ones.caseadesso partners with Anthropic to scale agent-based software engineering for customersCTOIf you roll out agent-based development through an external integrator, ask for concrete commitments on measurement and security.tool changeCodeRabbit offers cheaper NVIDIA model and application-level security scanningtech leadMeasure whether the cheaper model maintains review quality.caseArcade acquires Smithery, the MCP server registryopen sourceBitcoin developers request early access to AI modelssenior developerIf you work on such a project, expect AI-based vulnerability discovery.surveyIndia survey: 66% of AI and ML workers expect job cuts within 3–6 monthssenior developerThe survey finds engineers less afraid of layoffs, but it is still worth building agent skills.job marketEngineering managers return to the codebase as AI lowers the barriertech leadConsider how much time to spend coding with AI and what remains for leading the team.

← Issue 1 · Issue 3 →