Добірка вражень/Коллекция впечатлений

Elon Musk has weighed in on reports Amazon is addressing recent outages, including one related to AI-assisted coding.

The e-commerce giant held a mandatory meeting on Tuesday for a “deep dive” into multiple outages, including some as a result of the use of AI coding features, the Financial Times reported, citing internal briefs and emails. According to the outlet, Amazon said there was a “trend of incidents” in the past few months with a “high blast radius” and relating to “Gen-AI assisted changes,” as well as other variables.

Earlier this month, Amazon’s website and shopping app were down for some users, with more than 22,000 users reporting an issue, according to outage tracker Downdetector. Customers were unable to check out, view prices for goods, or access their account information. At the time, Amazon said the outage was a result of “a software code deployment.”

The report of the meeting drew the attention of tech experts, including Musk, who made his comments public when he responded to a post from Lukasz Olejnik, a cybersecurity consultant and visiting senior research fellow at Department of War Studies, King’s College London.

“Amazon is holding a mandatory meeting about AI breaking its systems,” Olejnik wrote.

“Proceed with caution,” Musk replied.

Dave Treadwell, Amazon’s senior vice president of e-commerce services, reportedly wrote in an email that the team’s weekly “This Week in Stores Tech” (TWiST) meeting would in part be used to implement additional guardrails on how AI is used by engineers, including requiring more senior engineers to sign off of AI-assisted changes made by junior and mid-level engineers.

“Folks, as you likely know, the availability of the site and related infrastructure has not been good recently,” Treadwell wrote in an internal email, the FT reported.

An Amazon spokesperson told Fortune that the TWiST meeting is a regular weekly operations meeting with a group of retail technology teams and leaders to review operational performance.

“As part of normal business, the meeting will include a review of the availability of our website and app as we focus on continual improvement,” the spokesperson said in a statement.

The company confirmed Amazon Web Services (AWS) was not involved in the incidents. Amazon said only one incident discussed was related to AI, but none involved AI-written code. Junior and mid-level engineers are also not required to have senior engineers sign off on AI-assisted changes, according to the company.

The outages and subsequent meeting has raised concerns from cybersecurity experts about risks associated with the rapid rollout of AI tools. Features like Amazon’s AI assistant Q can speed up the coding process, producing more code faster, but it may come at the risk of disrupting systems for how that code is written, checked, and deployed, making platforms more susceptible to outages, Olejnik told Fortune.

“I’m not making an argument against deployment of AI,” he said. “There isn’t any. It can’t be stopped. Everybody is going to deploy AI. It’s an argument against speed for its own sake or using AI for the sake of using AI.”

Late last year, Amazon began the process of laying off thousands of workers, citing desires to become more efficient and align the company culturally. Those layoffs have continued into this year, with the company reducing staff by a further 16,000 in January. Meanwhile, Amazon has continued pouring money into AI, projecting $200 billion in capex in 2026, an increase from $131 billion in 2025.

Musk, for his part, has previously said AI will bypass coding completely by the end of 2026.

Olejnik warned the transition from human-centered coding to AI-run systems too quickly could result in missing safety checks resulting in prolonged downtime or data loss that could result in “blowing up” a business due to irresponsible AI deployment.

When asked, he said he saw eye-to-eye with Musk regarding the level of attention AI deployment in tech should require.

“I agree with him,” Olejnik said. “AI brings a lot of opportunities, but there’s a middle ground between going to obsolescence due to not using AI, and blowing up businesses due to ill-judged deployments.”

Vibe-coding, in which developers simply tell AI what they need and sit back, is on the rise. Anthropic released Claude Code in February last year and passed $1 billion in annualized recurring revenue by November. Other tech giants have put out rival coding programs to try and snag those enterprise accounts. But while AI might code faster…it’s usually a lot messier: A report last year from software company CodeRabbit found that out of 470 pull requests (engineer speak for bug fixes or changes in code), AI code had 1.7 times more issues than human code.

“This sort of issue will become more prevalent because right now when a human operator acts, they do things with an understanding of the overall environment and knowledge of what they should and should not do,” according to research company Forrester Principal Analyst Brent Ellis. “An AI however will use whatever resources it has access to in order to try to achieve the goal it is given.”

Still, some analysts say it’s too early to conclude that AI-generated code will lead to more outages overall.

“The bar for AI code is the human error rate,” Constellation Research Principal Analyst Holger Mueller said. But, he noted, AI code could create more widespread security vulnerabilities and errors if all of the big cloud companies are drawing from the same AI coding platform.

Bottom line is Anthropic has introduced Code Review to Claude Code, a new feature that performs deep, multi-agent code reviews that catch bugs humans often miss, the company said.

Introduced March 9, Code Review is available in a research preview stage for Claude for Teams and Claude for Enterprises customers. Dispatching agents on a pull request, Code Review dispatches a team of agents that look for bugs in parallel, verify bugs to filter out false positives, and rank bugs by severity, according to Anthropic. The result appears in the pull request as a single, high-signal overview comment, plus in-line comments for specific bugs. The average review takes around 20 minutes, Anthropic said.

Anthropic has been running Code Review internally for months. On large pull requests (more than 1,000 lines changed), 84% get findings, averaging 7.5 issues. On small pull requests of fewer than 50 lines, the rate of findings drops to 31%, averaging 0.5 issues. Anthropic has found that its engineers mostly agree with what Code Review surfaces, marking less than 1% of findings as incorrect.