Business, Deals & Funding
Claude Code Changelog
v2.1.142
Version 2.1.142 of Claude Code adds new flags to the `claude agents` command for configuring background sessions (including directory, settings, MCP config, plugin directory, permission mode, model, effort, and a flag to skip permissions). Fast mode now defaults to Opus 4.7 instead of 4.6, with an environment variable override to pin to the older version. Plugin handling is improved so plugins with a root-level SKILL.md and no skills subdirectory are surfaced as a skill, and the plugin detail pane has been updated.
Why it matters
This is a solid incremental release focused on improving the agents workflow with more granular configuration flags, which should make background session management much more flexible. The model upgrade to Opus 4.7 for fast mode is a welcome default improvement, and the plugin skill surfacing change is a nice quality-of-life enhancement. The changelog entry appears truncated, so there may be additional changes not captured here.
NY Times
Nvidia’s Future in China Remains Unclear After Trump-Xi Summit
The article discusses the uncertain future of Nvidia's business in China following a Trump-Xi summit, highlighting how Chinese companies are increasingly shifting to domestic chipmakers such as Huawei as part of a broader effort to reduce reliance on Western technologies.
Why it matters
This reflects the ongoing and deepening tech decoupling between the US and China. The shift toward domestic chip alternatives like Huawei signals that US export controls, while intended to limit China's tech capabilities, may be accelerating China's push for self-sufficiency. For Nvidia, losing access to the massive Chinese market could significantly impact revenue, while China's domestic chip ecosystem, though still behind in some areas, stands to benefit from increased investment and demand.…
TechCrunch AI
What the jury will actually decide in the case of Elon Musk vs. Sam Altman
The article explains the specific legal questions that jurors will decide in Elon Musk's lawsuit against OpenAI, Sam Altman, Greg Brockman, and Microsoft. The jury is considering three claims: breach of charitable trust (whether OpenAI violated agreements about using Musk's donations for charitable purposes), unjust enrichment (whether defendants used Musk's donations to enrich themselves through OpenAI's for-profit arm), and aiding and abetting breach of charitable trust (whether Microsoft knowingly contributed to violating Musk's donation conditions). OpenAI's defense rests on three arguments: statute of limitations (that harms occurred before legally relevant dates), unreasonable delay (Musk waited too long to file), and unclean hands (Musk's own conduct was unconscionable). Key evidence includes that no witness could describe specific restrictions Musk placed on his donations, that…
Why it matters
This is a well-structured legal explainer that cuts through the noise surrounding this high-profile case to focus on what actually matters legally. The narrow framing of the jury's questions reveals that despite the dramatic narrative of feuding tech billionaires, the case hinges on relatively technical questions about charitable donation restrictions and timing. OpenAI appears to have a strong defense on multiple fronts — particularly the forensic accountant testimony about donations being spe…
The Verge AI
Closing time
The article covers the closing arguments in the Musk v. Altman trial, reported by Elizabeth Lopatto of The Verge. Musk's lawyer Steven Molo performed poorly, stumbling over words, misidentifying a co-defendant as 'Greg Altman,' and incorrectly claiming Musk wasn't seeking money before being corrected by the judge. OpenAI's lawyers countered effectively: Sarah Eddy organized the company's extensive evidence chronologically and delivered a notable line about Musk — 'Even the mother of his children can't back his story.' Co-counsel William Savitt highlighted the many times Musk claimed not to recall critical details and questioned how a sophisticated businessman could fail to understand a four-page term sheet from OpenAI. The article characterizes the closing as a 'demolition derby' and notes that while the trial produced considerable gossip, Musk's side offered little evidence supporting…
Why it matters
The article is clearly written from a perspective that views Musk's legal team as having performed disastrously in closing arguments, and the reporter makes little effort to hide her assessment that the defense dominated. The characterization as a 'demolition derby' and the detailed cataloging of Molo's errors versus the organized effectiveness of OpenAI's lawyers suggests the author believes Musk's case is weak. While the reporting appears factual in describing courtroom events, the framing he…
TechCrunch AI
Elon Musk’s SpaceXAI has been bleeding staff since its merger
More than 50 researchers and engineers have left Elon Musk's newly merged SpaceXAI since February 2026, when SpaceX acquired xAI. Key departures span coding, world models, and Grok voice teams, with at least 11 employees defecting to Meta and seven joining Mira Murati's Thinking Machines Lab. The pre-training team has been particularly hard hit following the exit of team lead Juntang Zhuang, raising concerns about the company's commitment to developing leading AI models. Reported causes include Musk's culture of extreme work and unrealistic deadlines, as well as reduced retention incentives as employees anticipate liquidity from SpaceX's expected IPO.
Why it matters
This article paints a concerning but unsurprising picture of what happens when Musk's notoriously demanding management style collides with the highly competitive AI talent market. The loss of the pre-training team is particularly significant — pre-training is foundational to building competitive AI models, and losing that expertise could set SpaceXAI back considerably. The merger of SpaceX and xAI always seemed more about consolidating Musk's empire than creating genuine synergies, and the tale…
The Verge AI
Behold, the Elon Musk jackass trophy
During the Musk v. Altman trial, OpenAI's legal team presented a trophy inscribed 'Never stop being a jackass' that OpenAI employees had purchased for AI safety researcher Josh Achiam. The trophy commemorates an incident where Elon Musk allegedly called Achiam a jackass after the researcher questioned Musk's desire to race ahead of Google in AI development. Altman's team used this to undermine Musk's current portrayal of his lawsuit as motivated by AI safety concerns, suggesting safety wasn't a priority for him at the time. Musk denied calling Achiam a jackass during his testimony, claiming he may have said 'Don't be a jackass.' Judge Yvonne Gonzalez Rogers ruled the jury could not see the trophy unless Musk's team gave reason to introduce it, so jurors only heard about the incident.
Why it matters
This is a delightfully theatrical courtroom moment that cuts to the heart of the Musk v. Altman case. The trophy is a clever piece of evidence because it serves as a tangible artifact contradicting Musk's narrative that he's motivated by AI safety concerns. If Musk indeed dismissed an AI safety researcher as a 'jackass' for raising legitimate concerns about racing ahead recklessly, it significantly undermines his current legal posture. The judge's decision to keep the trophy from the jury unles…
TechCrunch AI
OpenAI says Codex is coming to your phone
OpenAI has integrated its Codex coding tool into the ChatGPT mobile app on iOS and Android, allowing users to monitor and manage their development workflows remotely from their phones. The update, currently in preview and available to all plans, lets users work across threads, review outputs, approve commands, change models, and start new tasks. This follows recent additions including background desktop operation and a Chrome extension for live browser sessions. The move comes amid intense competition with Anthropic's Claude Code, which released a similar 'Remote Control' feature in February.
Why it matters
This is a logical and meaningful step in making AI-powered coding tools more accessible and practical. The ability to monitor and manage coding workflows from a phone addresses a real need for developers who want to stay connected to autonomous coding agents without being tethered to their desktops. The competitive dynamic between OpenAI and Anthropic is clearly driving rapid feature development, which benefits users. However, the pace of these releases raises questions about whether the tools…
OpenAI
Sea's View on the Future of Agentic Software Development with Codex
Sea Limited, a Singapore-founded global tech company operating across e-commerce, digital entertainment, and financial services in Southeast Asia, is rolling out OpenAI's Codex across its entire development organization. According to Co-Founder and Shopee CPO David Chen, 87% of Codex users are weekly active users. Chen describes Codex not as a marginal productivity tool but as a 'structural multiplier' that helps engineers manage large-scale systemic complexity across fragmented, hyper-localized markets. He highlights Codex's deep contextual awareness of large codebases and its ability to act as a 'localized knowledge engine,' reducing time spent navigating unfamiliar microservices and shifting cognitive load toward higher-level tasks like architectural design and product innovation. Strong internal usage is reported across code understanding, debugging, and feature development. Sea has…
Why it matters
This article reads as a polished corporate partnership announcement and promotional piece rather than genuine journalism or independent analysis. It's published on OpenAI's own website and follows a predictable format: softball questions, glowing answers, and impressive-sounding metrics (87% weekly active users) presented without context or independent verification. There's no discussion of challenges, limitations, failed experiments, or developer pushback. The language is saturated with corpor…
From X/Twitter
- RoundtableSpace suggests asking Claude to map your entire app's architecture into one HTML page and one JSON file — the HTML is for you, the JSON is for the next agent.
- Alfie Carter spent 55 hours documenting exactly how GTM engineers run 125 skills across 6 layers — covering lead enrichment, outbound sequencing, LinkedIn content, and programmatic SEO.
- Useful API trick from ClaudeDevs: pre-warm the prompt cache by sending your system prompt before the user prompt to cut time-to-first-token on longer calls.
- Two senior staff engineers at Airbnb gave a live lecture on how they actually build with agents in 2026 — after shipping one of the most ambitious LLM-agent migrations in production.
- 90% of people hitting Claude's usage limit are forcing it to re-read their entire context every message, says Suryanshti — and it's fixable.
- BCG data shows countries with the highest GenAI adoption also report the highest fears around job loss — proximity to AI-enabled workflows is accelerating anxiety, not reducing it.
From Reddit/HN/YC
- [Hacker News] Rosetta Check 2.0 scans your Pro Tools, Logic, and Photoshop plugins for lingering Intel builds.
- [Hacker News] Someone built vi for the 6502 and it works exactly like you'd hope.
- [Hacker News] Anthropic publishes two scenarios for global AI leadership by 2028 — and neither is comfortable.
- [Hacker News] A zero-day exploit completely defeats default BitLocker protections on Windows 11.
- [Hacker News] A script that turns a bare VPS into an operational fortress in 15 minutes and one command.
- [Hacker News] Cerebras bet on dinner plate-sized AI accelerators. Today it's worth $66B.