Friday Jul 17
FLOPSxAI

Grok Build Quietly Uploaded Your Code

17JUL
GROK BUILDYOUR CODE

xAI's coding tool secretly copied your work to its servers. Grok Build uploaded entire Git repos, including secrets, to Google Cloud. Musk promised a fix, but there's still no verified timeline.

A researcher's wire-level analysis found the uploads ran regardless of privacy settings. The tool sent about 27,800 times more data than any coding task needed.

The story hit Hacker News on July 14, forcing a public response. Musk promised to delete all collected user data. xAI has not said how many users were affected or for how long.

xAI then open sourced the entire Grok Build codebase, hoping to rebuild trust. The upload code is reportedly still present, with no independent proof the deletion happened.

full brief & sources

⚡ Why this matters

  • Developer tools that touch your codebase carry real trust risk if they misbehave.
  • A privacy toggle that does nothing undermines every other privacy claim a vendor makes.
  • Committed secrets in uploaded repos means API keys and credentials may already be exposed.

🔍 What happened

  • A security researcher published a wire-level analysis on July 12, 2026.
  • Grok Build, xAI's coding CLI tool, uploaded full tracked Git repositories to a Google Cloud Storage bucket.
  • The bucket was named grok-code-session-traces, reachable without user consent or disclosure.
  • The privacy toggle marketed as 'Improve the model' had no effect on the uploads.
  • The story hit Hacker News front page on July 14, 2026.
  • xAI open-sourced the Grok Build codebase on July 16, days after the backlash.

💬 Smart takes

  • The Register: Musk promised a purge of previously uploaded user data after the story broke.
  • Skeptic: xAI has given no user count, no data volume, no verification method, and no deletion timeline.

🧭 Where this goes

  1. LikelyxAI faces a formal regulatory inquiry into the undisclosed data collection.
  2. Likelydevelopers audit other AI coding tools for similar covert uploads.
  3. Possiblea class action lawsuit follows if committed secrets are shown to have leaked.
  4. Wild CardxAI publishes a third-party audit proving full deletion, resetting trust quickly.

🥄 The Spoon Take

A coding tool that uploads your repo without telling you is not a bug. It is a design choice someone shipped anyway. Open-sourcing the code after getting caught does not answer the real question: how much of your data is already sitting in that bucket.

🤔 Pushback

Musk's team moved fast to open source and respond publicly, which is more transparency than most vendors offer after a breach.

·The Register - Musk promises purge·Simon Willison - Grok Build open source·TechTimes - Repos, secrets stay in code
Friday Jul 10
PRODUCTxAI

Grok 4.5 Launches At A Steep Discount

10JUL
GROK 4.5PRICE CUT

xAI just started a frontier-model price war. Grok 4.5 prices in more than 60% below Opus 4.8 and GPT-5.5. It's the first xAI model trained on real Cursor coding sessions.

xAI shipped Grok 4.5 on July 8. It's built for exactly one job: coding and agent work.

Cache hits cut input cost to 50 cents per million tokens. The context window shrank to 500,000 tokens, down from 1 million. It still ranks fourth on the Artificial Analysis Intelligence Index, above every open model.

Musk called it Opus-class the day it shipped. Cheap and fast is the new baseline now, not a selling point.

full brief & sources

⚡ Why this matters

  • Frontier models are converging on capability, so price and specialization are becoming the real battleground.
  • Training a flagship model on real coding-tool session data, not just scraped code, is a new recipe.

🔍 What happened

  • xAI released Grok 4.5 on July 8, its first model built specifically for coding and agentic tasks.
  • API pricing: $2 per million input tokens, $6 per million output tokens; cached input drops to $0.50.
  • That's over 60% cheaper than Anthropic's Opus 4.8 and OpenAI's GPT-5.5.
  • Grok 4.5 trained partly on real Cursor developer session data, not synthetic benchmarks.
  • It ranks fourth on the Artificial Analysis Intelligence Index, above all open-weight and Gemini models.
  • Context window is 500k tokens, down from Grok 4.3's 1 million; not yet available in the EU.

💬 Smart takes

  • Elon Musk: called Grok 4.5 "Opus-class" on launch.
  • Axios: framed the release as a scoop on xAI's push into developer tools.
  • Skeptic: a shrunk context window, 500k tokens down from 1 million, is a real regression dressed up as a pricing win.

🧭 Where this goes

  1. LikelyAnthropic and OpenAI respond with cheaper, task-specific coding tiers within weeks.
  2. LikelyCursor session data becomes a contested resource other labs try to license or replicate.
  3. PossiblexAI's EU delay becomes a pattern as it prioritizes US enterprise deals first.
  4. Wild CardGrok 4.5's price undercut is steep enough that it becomes the default model inside Cursor itself.

🥄 The Spoon Take

Model quality is converging, so price just became the weapon. xAI trained on the exact coding sessions rivals want, then priced it to force a reaction. Watch what Anthropic and OpenAI do to their coding tiers in the next month, that's the real scoreboard.

🤔 Pushback

A shrunk context window and no EU access are real limits; cheap and fast doesn't help if the job needs long context or lands outside the US.

·x.ai - Introducing Grok 4.5·Axios - Grok 4.5 scoop·Kingy.ai - Grok 4.5 benchmarks
Monday Jul 6
COOL TOOLSxAI

xAI Ships One-Model Voice Agents

6JUL
1 MODEL

Voice agents used to need three stitched models. xAI now turns plain text into a live phone agent in two minutes. It beat Google and OpenAI on the speech benchmark using one model, not three.

Under the hood: live call handling, real-time lookups, and safety checks ship together already, no extra integration work.

It answers in under 700 milliseconds, faster than pipelines that turn sound into words and back. It also keeps laughs, whispers, and sighs that a written middle step would erase.

Pricing lands at five cents per minute, with cloning and eighty-plus voices bundled in. Access is still limited, and some builders are hitting errors.

full brief & sources

⚡ Why this matters

  • Voice UX has been the weak link in AI agents. Latency and stitched pipelines make them feel robotic.
  • One speech-to-speech model removes two translation steps, which is usually where the lag comes from.
  • The pricing undercuts building an in-house voice stack from scratch.

🔍 What happened

  • xAI launched Voice Agent Builder in beta on July 1, 2026.
  • No-code: describe a phone call in plain language, get a live agent in under two minutes.
  • Runs on one speech-to-speech model instead of three stitched APIs.
  • Responds in under 700 milliseconds. Scored 67.3% on the tau-voice Bench, beating Gemini 3.1 Flash Live at 43.8% and GPT Realtime 1.5 at 35.3%.
  • $0.05 per minute. 80-plus voices, voice cloning from two minutes of audio, 25-plus languages with mid-call switching.

💬 Smart takes

  • eesel AI review: developers who built with it are impressed by mid-conversation language switching and how fast a working agent comes together.
  • Skeptic: it's still beta. Several developers hit 403 access errors, and no one has published a fix for an agent acting on a misheard instruction.

🧭 Where this goes

  1. LikelyxAI opens broader access within weeks once the access errors get sorted.
  2. Likelyrival labs answer with their own single-model voice stack within a quarter.
  3. Possibleagencies use it to replace call-center phone trees, not just simple bots.
  4. Wild Cardthis becomes the default way small businesses build a phone line within a year.

🥄 The Spoon Take

The voice-agent race has been about who has the smartest model. xAI just made it about who has the fastest one. Sub-second, single-model voice is the unlock that finally makes AI phone agents feel less like a phone tree.

🤔 Pushback

A benchmark win on one leaderboard doesn't mean it holds up on messy real calls with accents, noise, and interruptions.

·xAI - Voice Agent Builder launch·eesel AI - Hands-on review·Basenor - Beta launch details
Tuesday Jun 30
STRATEGYxAI

Grok 4.5 Tests At Tesla And SpaceX

28JUN
1.5T PARAMSGROK 4.5SPACEX

Elon Musk put Grok 4.5 into private beta at Tesla and SpaceX first. It runs on a 1.5-trillion-parameter base, trained partly on Cursor coding data. Early evals put it near Anthropic's Opus.

Musk's own companies are the test lab. Tesla and SpaceX engineers run Grok 4.5 before the public sees it. Real workloads, real feedback, fast.

The Cursor data matters. SpaceX bought Cursor for $60 billion this month. Now its coding data trains Grok. xAI says it ships a freshly trained model every month.

Vertical integration is the bet. Own the coding tool, own the data, own the model. Watch whether monthly retrains hold quality or just chase headlines.

full brief & sources

⚡ Why this matters

  • xAI is using its own companies as a private testing ground before public release.
  • Grok 4.5 trains on Cursor data, tying the SpaceX-Cursor deal directly into model quality.
  • A monthly model cadence is aggressive versus rivals' quarterly cycles.

🔍 What happened

  • Jun 28: Musk says Grok 4.5 is in private beta at Tesla and SpaceX.
  • Built on xAI's 1.5-trillion-parameter V9 foundation model.
  • Cursor coding data added in supplemental training.
  • Early evals reportedly near or above Anthropic's Opus.
  • xAI plans freshly trained models every month this year.

💬 Smart takes

  • Elon Musk: Grok 4.5 early evals show performance close to, perhaps exceeding, Opus.
  • Skeptic: in-house evals from the CEO are marketing until independent benchmarks confirm them.

🧭 Where this goes

  1. LikelyGrok 4.5 hits public release within weeks of the beta.
  2. LikelyxAI leans on Cursor data as a coding-quality edge.
  3. Possiblemonthly retrains strain compute and quality control.
  4. Wild CardTesla ships a Grok-powered coding assistant to outside developers.

🥄 The Spoon Take

Musk is building a closed loop. Buy the coding tool, feed its data to the model, test it on your own engineers, ship monthly. If it works, xAI's data edge compounds. If the monthly pace breaks quality, it's just noise dressed as speed.

🤔 Pushback

CEO-reported evals and a private beta aren't a product. Until outsiders test Grok 4.5, near-Opus is a claim, not a result.

·Times of AI - Grok 4.5 private beta·TechTimes - Grok V9 model·xAI - News
Friday Jun 26
PRODUCTxAI

Grok Codes On Its Own

26JUN
RUNS UNATTENDEDGROKAUTONOMOUS

xAI gave Grok a hands-off mode. Type a goal in Grok Build, and the agent plans, writes, tests, and verifies code until the task is done. No human babysitting.

This is xAI's answer to Codex and Claude Code. The pitch: hand off a whole job, walk away, come back to finished work.

Under the hood it splits each job across three models, one per stage. Controls let you pause or check progress. It needs a paid SuperGrok or X Premium Plus subscription.

The bet across every lab is the same. Whoever can run longest without a person watching wins the developer.

full brief & sources

⚡ Why this matters

  • Long-running autonomous agents are the new coding battleground.
  • Grok joins Codex and Claude Code in the unattended-agent race.
  • Multi-model plan-build-verify pipelines are becoming the standard pattern.

🔍 What happened

  • Jun 22: xAI launched /goal in Grok Build.
  • Give one goal; the agent plans, executes, tests, and verifies until complete.
  • Controls: /goal status, pause, resume, clear.
  • Runs Composer 2.5 to plan and Grok Build 0.1 to implement, plus a verifier.
  • Needs a SuperGrok ($30/mo), SuperGrok Heavy ($300/mo), or X Premium Plus ($40/mo) plan.

💬 Smart takes

  • xAI: /goal handles 'larger implementation tasks' end to end with built-in verification.
  • Skeptic: 'runs until verified' is easy to claim, hard to trust on real codebases. Long runs burn tokens fast.

🧭 Where this goes

  1. Likelyevery coding agent ships a long-running autonomous mode this year.
  2. Likelyverification quality, not raw speed, becomes the selling point.
  3. Possibleagent-hours and token burn become the real cost debate for dev teams.
  4. Wild Carda team ships a production feature start to finish with zero human edits.

🥄 The Spoon Take

The coding agent race moved from autocomplete to autonomy. The question is no longer 'can it write code' but 'how long can it run alone before it breaks something.' Grok, Codex, and Claude Code are all chasing the same prize: the agent you can leave running overnight. Verification is the moat now.

🤔 Pushback

Autonomous runs sound great until one quietly ships a bug at hour six. Trust, not capability, is the real blocker.

·xAI - Introducing /goal·MarkTechPost - Grok Build /goal
Thursday Jun 25
PRODUCTxAI

Grok Lands Free Inside Microsoft Office

25JUN
$0 VS $30 GROK WORD

xAI just undercut Microsoft inside Microsoft's own apps. Grok shipped free add-ins for Word, Excel, and PowerPoint with live web research. Copilot costs $30 a seat. Grok costs nothing through 2026.

Redmond's assistant runs $30 a head. xAI decided to match the feature set and charge zero.

The plug-ins sit in the same document panels and draft, rewrite, and fetch current results, and they cite their sources. There are no usage caps and no price tag planned for this calendar year.

This is a distribution grab funded out of pocket. The aim is not income. It is the user base that pays a rival every month.

full brief & sources

⚡ Why this matters

  • A free, capable rival inside Office attacks Microsoft's strongest AI distribution.
  • Free-with-no-limits is a war-chest move, not a business model.
  • It puts pricing pressure on every paid Office AI add-on.

🔍 What happened

  • Jun 18: xAI launched Grok add-ins for Word, Excel, and PowerPoint.
  • The add-ins read the whole document and generate or rewrite text.
  • They run live web search through xAI, Brave, and Bing with source links.
  • xAI says the add-ins are free with no token limits.
  • Microsoft 365 Copilot still costs $30 per user per month.

💬 Smart takes

  • Windows News: xAI planted Grok inside Word's side panel, and it is free.
  • xAI: no premium tiers and no plans to charge for the add-in during 2026.
  • Skeptic: free until it isn't; once Grok has the users, the 2027 price tag is the real product.

🧭 Where this goes

  1. LikelyMicrosoft tightens add-in rules or bundles more Copilot value to defend the seat.
  2. Likelyother AI vendors ship free Office add-ins to ride the same channel.
  3. Possibleenterprises block third-party add-ins on data-governance grounds.
  4. Wild CardGrok's free Office push forces Copilot to drop its $30 price within a year.

🥄 The Spoon Take

Giving the product away inside a rival's app is an attention play, not a money play. xAI is buying its way onto millions of screens while Microsoft charges for the same seat. The bet is that distribution now beats margin. If it works, every paid AI add-on has to explain why it costs anything.

🤔 Pushback

Free third-party add-ins die fast in the enterprise. IT blocks what it can't govern, and a side-panel bot is easy to block.

·Windows News - Grok in Word·Windows News - Office add-ins·Blockchain.news - Grok Word add-in
Tuesday Jun 23
COOL TOOLSxAI

Grok Video Undercuts Sora On Price

16JUN
GROK VIDEO14c A SEC

xAI's new video model topped the leaderboard and slashed the price. Grok Imagine Video 1.5 outranks Sora 2, Veo 3.1, and Kling. It runs at 14 cents a second.

It shipped straight to everyone, no waitlist. You can use it right now in the app, on the web, and through the developer API.

The cut is the real shift, roughly 86% cheaper than the market leader. A clip that cost five dollars now costs well under one. That moves AI video from studio budgets to anyone's pocket.

Cheap plus good usually forces everyone else to drop prices too. Expect rivals to answer within weeks. For creators, the math on making short video just flipped in their favor.

full brief & sources

⚡ Why this matters

  • Top of the video arena and the cheapest at once. That combo is rare.
  • Price is the lever. 14 cents a second resets what creators expect to pay.
  • xAI shipped it to consumers, not a waitlist. API, web, and mobile on day one.

🔍 What happened

  • Jun 16 - xAI launched Grok Imagine Video 1.5 into general availability.
  • Live across the API, grok.com, and the iOS and Android apps.
  • Tops the Image-to-Video Arena leaderboard with a +52 Elo gain over 1.0.
  • Outranks Sora 2, Veo 3.1, and Kling on that board.
  • Priced at 14 cents per second at 720p, about 86% below Sora 2 Pro's 50 cents.

💬 Smart takes

  • xAI: frames 1.5 as a clear leaderboard win at a fraction of rival pricing.
  • BuildFastWithAI: flagged the price cut as the headline, not the Elo score.
  • Skeptic: arena Elo is not real-world quality. Cheap clips can still look cheap.

🧭 Where this goes

  1. Likelyrivals cut video prices within a quarter to defend share.
  2. Likelycreators test Grok for short social clips where cost matters most.
  3. Possiblea per-second price war pushes AI video toward commodity pricing.
  4. Wild Carda major app embeds Grok video as its default generator by year end.

🥄 The Spoon Take

Video models stopped competing only on quality. Now they compete on price. xAI took the leaderboard and undercut the field by most of an order of magnitude. When the best clip is also the cheapest, the moat moves from the model to distribution. Whoever puts video in front of the most people wins.

🤔 Pushback

Leaderboard rank and headline pricing are easy to game. Real production quality, rights, and rate limits decide what creators actually ship.

·BuildFastWithAI - Grok video 1.5·LaunchToolsAI - June 2026 tools
Wednesday Jun 17
PRODUCTxAI

Grok Gets A Plugin Marketplace

17JUN
6 AT LAUNCHGROKPLUGINS

xAI wants Grok to be a coding platform, not just a chatbot. It launched a plugin marketplace for Grok Build, its coding agent. MongoDB, Vercel, Sentry, and Cloudflare are live at launch.

The plugin, not the model, is the product now. Whoever owns the extension layer owns the developer's day.

The real story is where xAI got its catalog. It adopted Claude Code's plugin format wholesale, including the 224,000-star Superpowers pack. Years of ecosystem work, skipped in one move.

That makes the standard portable. An add-on built for one tool may soon run on all of them. Good for developers. Bad for any lab betting on lock-in.

full brief & sources

⚡ Why this matters

  • The coding-agent fight is shifting from model quality to ecosystem depth.
  • Plugins, skills, and MCP servers are becoming the real switching cost.
  • xAI adopting Claude Code's plugin format hints at an emerging shared standard.

🔍 What happened

  • Jun 11: xAI launched the Grok Build Plugin Marketplace in beta.
  • Six plugins at launch: MongoDB, Vercel, Sentry, Chrome DevTools, Cloudflare, Superpowers.
  • A plugin bundles skills, slash commands, agents, hooks, MCP servers, and LSPs.
  • Grok Build is xAI's terminal coding agent, in beta since mid-May.
  • Each remote plugin is pinned to a commit and verified at install.
  • Live for SuperGrok and X Premium Plus subscribers, no extra fee yet.

💬 Smart takes

  • xAI: developers can browse, install, and update plugins from the terminal.
  • MarkTechPost: xAI is pushing deeper into AI-assisted developer tooling.
  • Skeptic: a six-plugin beta is tiny next to Claude Code and Codex. Catalog depth decides this, not the storefront.

🧭 Where this goes

  1. Likelythe plugin catalog grows fast as xAI courts infrastructure vendors.
  2. Likelyportable plugin formats become normal across coding agents this year.
  3. Possiblea cross-agent standard emerges so one plugin runs on Grok, Codex, and Claude Code.
  4. Possiblesecurity reviews of third-party plugins become a real enterprise concern.
  5. Wild Carda popular plugin author monetizes directly and starts an agent-plugin economy.

🥄 The Spoon Take

The coding agent is becoming an operating system. Whoever owns the plugin layer owns the developer's habits. xAI skipped years of ecosystem building by adopting Claude Code's format wholesale. That is flattering to Anthropic and dangerous to it. Standards that travel are good for developers and bad for lock-in.

🤔 Pushback

A storefront with six plugins is a press release, not an ecosystem. Grok Build needs thousands of quality plugins before this matters.

·xAI - Grok plugin marketplace·MarkTechPost - Grok plugins ship·Crypto Briefing - MongoDB plugin live
Tuesday May 26
PRODUCTxAI

The Coding Agent Wars Hit Four

25MAY
CLAUDECODEXAGY (G)GROKNEW!FOUR-WAY WAR!CODING AGENTS BAKE-OFF

xAI opens Grok Build, its coding agent, to all paying subscribers. Four coding agents now compete: Anthropic's Claude Code, OpenAI's Codex, Google's agy, and xAI's Grok Build.

Single-vendor lock-in just got structurally weaker. Configs port across vendors. Switching is a 2-week decision, not a 2-year one.

Differentiation moves up the stack to IDE, plugins, and ecosystem. The CLI is now table stakes. The IDE is where the wars get fought next. Independent tools (Cline, Aider, Continue.dev) face acqui-hire pressure within 12 months.

Run a 30-day bake-off this quarter on a real codebase before standardizing. Pick winners on per-PR token cost and completion rate, not vendor preference.

full brief & sources

⚡ Why this matters

  • Coding-agent CLIs went from one vendor to four in five months.
  • Single-vendor lock-in is now a real procurement risk, not a hypothetical.
  • Switching costs are low. Configs port across vendors.

🔍 What happened

  • May 25, 2026. xAI opens Grok Build to all SuperGrok and X Premium Plus subscribers (early beta).
  • Initial gated launch was May 14 to SuperGrok Heavy.
  • Elon Musk personally promoted it on X with a "Tips" guide on May 21.
  • Four-vendor landscape: Claude Code (Anthropic), Codex CLI (OpenAI), agy (Google, replacing Gemini CLI), Grok Build (xAI).
  • All four are terminal-native. All read CLAUDE.md / AGENT.md-style project configs.

💬 Smart takes

  • xAI product page: "Terminal-native, agentic command-line interface built for professional software engineering."
  • DevOps.com: "Grok joins Claude Code, Codex, Antigravity in the coding agent race."
  • Skeptic: Grok 4 lags Claude Opus 4.7 and GPT-5.5 on coding benchmarks. $50/month X Premium Plus paywall narrows the market vs Claude Code's free tier.

🧭 Where this goes

  1. First published bake-offs (GitHub, Stack Overflow, JetBrains) land by mid-Q3.
  2. AGENT.md / CLAUDE.md becomes the de-facto portable config spec by Q3.
  3. Independent coding-agent tools (Cline, Aider, Continue.dev, Sweep.dev) face acqui-hire pressure.
  4. IDE-level competition becomes the next axis (Cursor vs Antigravity Desktop vs Codex IDE vs Claude Code IDE).

🎯 Implication

  • For engineering leaders: stop assuming Claude Code is the default. Run a 30-day head-to-head this quarter on a real codebase. Measure per-PR token cost and completion rate.
  • For execs: add coding-agent CLI choice to your vendor concentration risk model.
·xAI - Introducing Grok Build (product page)·eWeek - Elon Musk's xAI launches Grok Build, a coding agent·DevOps.com - xAI enters the coding agent race with Grok Build