Author: Amitabh Sarkar

Bun, the JavaScript runtime Anthropic acquired in December 2025, has been rewritten from Zig to Rust largely by Claude Fable 5. The AI-assisted rewrite produced 1,009,257 lines of Rust at an estimated API cost of $165,000, with 99.8% of tests passing on Linux x64. Bun founder Jarred Sumner announced the Rust rewrite on July 8, 2026 on the official Bun blog, and the announcement carried a second surprise: Anthropic acquired Bun in December 2025, a deal not widely reported at the time. The rewrite converted Bun’s 535,496-line Zig codebase into more than a million lines of Rust using a pre-release…

Grok 4.5 is SpaceXAI’s newest and most capable AI model, launched July 8, 2026. Priced at $2 per million input tokens and $6 output, it delivers coding performance competitive with Claude Opus 4.8 at roughly one-fifth the cost, though independent benchmarks rank it fourth overall behind Claude Fable 5 and GPT-5.5. SpaceXAI made Grok 4.5 public on July 9, 2026, one day after announcing it on July 8, positioning the model directly against Anthropic’s Claude Opus line. Elon Musk called it “an Opus-class model, but faster, more token-efficient and lower cost,” as reported by TechCrunch. It is the company’s first…

GPT-Live is OpenAI’s new full-duplex voice model that replaces the previous ChatGPT Voice feature, launched July 8, 2026. Unlike the old turn-based system, GPT-Live listens and speaks simultaneously — acknowledging you with “mhmm” and “got it” mid-sentence, interrupting when needed, and routing complex tasks to GPT-5.5 in the background without dropping the conversation. GPT-Live-1 is now the default for paid ChatGPT users. What Changed Launch date: July 8, 2026 Full-duplex: AI listens and speaks simultaneously — no more robotic turn-taking Natural acknowledgments: says “mhmm” and “got it” while you’re still talking Smart delegation: routes complex queries to GPT-5.5, returns answers…

Update — August 25, 2026: Two months after launch, Fable 5 accounts for just 11% of what enterprises spend on Anthropic products, according to Ramp data reported by the Financial Times. Claude Opus 5 — launched July 24 at half the price — has already overtaken it in dollar spend. See the full breakdown below. Update — July 19, 2026: The Fable 5 free-access period ends tonight at 11:59:59 PM PT. Starting July 20, continued Fable 5 use on paid plans requires separate prepaid usage credits. All other Claude models remain within standard plan limits. Claude Fable 5’s free subscriber…

🏻 Update — August 13, 2026 — Gemini 3.7 Flash launches, tops FrontierCode: Google launched Gemini 3.7 Flash on August 13, 2026, three weeks after Gemini 3.6 Flash. The model scores 43.6% on FrontierCode 1.1 Main — above Claude Sonnet 5’s 42.7% — and launches at exactly half the price of its predecessor: $0.75/M input + $3.75/M output through December 31, 2026 (rises to $1.50/$7.50 on January 1, 2027). Context window: 1 million tokens; max output: 64,000 tokens. Available now in Gemini Spark. Sources: Google Blog · SiliconAngle. Full analysis below ↓ 🔴 Update — August 5, 2026 — Leadership…

A GitHub AI agent security flaw discovered by Noma Labs lets an unauthenticated attacker steal private repository data with zero credentials and zero coding skill — just a crafted GitHub Issue posted to any public repo in the same organization. GitHub’s response was a documentation update, not a code fix, because prompt injection in AI agents cannot be fully patched in software. Key Facts Vulnerability named GitLost, discovered by Sasi Levi at Noma Labs, disclosed July 6, 2026 Affects GitHub Agentic Workflows — the new feature that pairs GitHub Actions with an AI agent backed by Claude or GitHub Copilot…

Updated September 2026: Anthropic has also opened self-hosted Claude Code environments to public beta for Team and Enterprise plans (see the new section below). Claude for Government has moved from its July 2026 public beta to general availability. All US federal, state, and local government agencies can now access the service — it is no longer limited to invite-only participants. Federal agencies procure through Carahsoft via the GSA Schedule or by contacting Anthropic directly at claude.com/solutions/government. Claude for Government is Anthropic’s FedRAMP High authorized product that brings Claude Code and Claude Cowork to US government agencies. Originally launched as an…

DeepSeek, the Chinese AI startup that upended the global AI race in early 2026, is now developing its own AI inference chip to reduce dependence on NVIDIA and Huawei hardware. Reported by Reuters on July 7–8, 2026, and confirmed by Bloomberg, the effort is still at an early stage — but DeepSeek is already in talks with manufacturing partners and quietly hiring chip engineers to staff the initiative. What Reuters and Bloomberg Reported According to Reuters, DeepSeek is developing its own chip specifically for inference — the process of running a trained AI model to generate responses — rather than…

ChatGPT went down for many users on the morning of July 12, 2026, with iOS users hitting 403 errors when trying to load or continue conversations. The disruption began around 5:29 AM ET, OpenAI confirmed active issues on its status page, and the service was reported operational again by 9:46 AM UTC — a short but visible outage during GPT-5.6 launch week. ChatGPT is down for a portion of users right now because of a 403 error affecting iOS conversations, first reported around 5:29 AM ET on July 12, 2026. OpenAI’s status page confirmed it was investigating, and monitoring service…

Uber burned through its entire 2026 AI tooling budget in just four months, the company’s CTO confirmed, and has now imposed a $1,500-per-employee monthly cap on each agentic coding tool — a hard limit that applies separately to Claude Code and Cursor. The reversal marks one of the most visible corporate pullbacks from the “use AI as much as possible” posture that dominated enterprise tech in late 2025. What Uber’s AI Spending Cap Actually Means According to reporting from Bloomberg and TechCrunch, the $1,500 monthly cap applies per tool, not per employee across all tools combined. An engineer using both…