Living in Grok all day so you don’t have to. Tips, drops, and the occasional hot take.
Colossus 3
Joined August 2025
- Tweets1.4K
- Following9
- Followers35
- Likes187
Grokverse retweeted
Grok Bot 0.75.8 is out. Fewer approval prompts, same control.
→ Bots can suggest changes to your Auto-review rules. You confirm on a card, nothing changes before that
→ Auto-review double-checks an action before asking you, so safe steps stop interrupting
→ New connector tools (like X) work right away, no reconnecting
→ When background work ends, the bot runs the follow-ups you asked for and stops there
A huge update that makes agents much less tiring to use.
Grokverse retweeted
AI didn't break the internet. It sent the bill.
Millions of attack attempts. No organization says it was breached. Two vendors withdrew their claims.
Inside: 8 AI predictions that missed, and a checklist if you run a site.
Has your host patched yet?
Grokverse retweeted
Weekly AI Recap, Sunday edition. Here are the seven biggest breakthroughs from the week.
1. Terminal X Eval Agent → 88.6% match with senior investor judgment, and 25% more mandate deviations caught on GPT-6.1 Sol. Every graded output is scored against a 54-criterion rubric.
2. Nemotron 3.5 Bolt 74B → 74B total parameters with 9B active, shipped as native 2-bit (IQ2) weights that fit in 24GB. A 74B dense model must keep all 74B parameters resident in memory.
3. Mistral Voxtral 3 → speech split into three layers: 3 transcription models, 1 text-to-speech model, and a LiveKit-based voice agent SDK that wires them together.
4. Google Cloud Agent Compute Engine → one hardware-isolated microVM per AI agent, with no credentials inside it.
5. Nemotron 3.5 Bolt E4B 4GB → an open small language model for local voice agents, game characters and robots. NVIDIA starts from Gemma 4 E4B and changes its behavior through post-training rather than a new base.
6. Kimi K3.1 Pro → a post-trained successor to Kimi K3 with a 1,048,576-token context. The story is post-training, not a new base release.
7. Agent skills donated to the Agentic AI Foundation → procedures no longer need rewriting for each platform, so a skill can survive the next release instead of depending on one company's roadmap.
Bookmark this recap to catch up on the week in 60 seconds.
Grokverse retweeted
Insane Grok @bot upgrade!
Finally, you can connect Grok Bot to different agent harnesses and run tasks through their native runtimes.
Devs just open-sourced HarnessRouter, the infrastructure layer that lets Grok Bot run tasks across:
- Codex
- Hermes
- Claude Code
- DeepSeek Harness
- System One, powered by Jev
- Pi, and 20 more
GitHub repo: github.com/HarnessRouter/har…
(don’t forget to star it ⭐)
With HarnessRouter, each external specialist gets to use its own harness, model, tools, skills, and agent loop while Grok coordinates the project from one conversation.
I tested this by asking Grok Bot to turn a support-ticket CSV into an interactive dashboard.
Grok Bot assigned the analysis to DeepSeek Harness with Claude Sonnet 5.5, passed the output to Codex with GPT-5.5, and checked the finished dashboard in its browser.
This differs from simply creating more Bots. These workers are not additional Grok Bots, and Grok Bot itself does not switch harnesses.
Instead, it remains the coordinator while the delegated tasks run inside separate HarnessRouter sessions with the harness of your choice.
So for another task, you can use Claude Code, Hermes, Pi, etc., without requiring a separate integration for task execution, sessions, streaming, and generated files.
HarnessRouter standardizes all those operations through the Unified Harness Protocol (UHP).
UHP handles sessions, streaming, files, cancellation, and failures through the same interface, regardless of which harness runs the task.
I shared the GitHub repo above.
And if you want to dive deeper into Grok Bot, I wrote about it in full in a masterclass, including the shared computer, memory, skills, routines, and the macOS setup.
Read it below.
Grokverse retweeted
Coding agents invent imports, packages, and method signatures that don't exist in your repo, and the errors surface only at runtime. This rule forces verification before any new call is written.
Paste this into your CLAUDE md or Cursor rules:
1. Before writing any import, check the project's dependency manifest, lockfile, or requirements file and the existing imports in the repo. If the package is missing, stop and list it as a proposed addition instead of guessing.
2. For any library function or method you have not seen used in this repo, read the installed source or local type definitions first. Do not rely on memory of the API version.
3. Make the smallest diff that satisfies the request. Do not rename, reformat, or refactor unrelated code. List each changed file with a one-line reason.
4. If a verification step fails or an import cannot be resolved, report the exact error and propose one fix. Do not make more than two guesses without asking.
Bookmark this rule for your next agent session.
Grokverse retweeted
Five breakout AI repos climbing the charts this week and the specific bottlenecks each one unblocks:
1. anthropics/claude-code
→ Official CLI adding a retry watchdog that stops unattended runs at distant usage-limit resets instead of hanging, plus OpenTelemetry tracing for observability
2. tamaratran/fast-jev-compaction
→ ~254 stars/day TypeScript plugin that swaps Claude Code's compaction summary for Jev scoring, cutting context bloat by evaluating every tool call in one fast request
3. rehan-remade/universal-modder
→ ~471 stars/day Python toolkit pointing Claude Code at any PC game for recon, reverse engineering, and asset generation via fal MCP
4. All-Hands-AI/OpenHands
→ v1.27.0 ships OpenCode and Pi as built-in ACP agents with full e2e verification on DeepSeek, replacing managed agent platforms
5. browser-use/jev-ultrafast
→ ~756 stars/day Python web agent claiming fastest and cheapest browser automation, an alternative to hosted scraping services
Bookmark this stack for your next build.
Grokverse retweeted
Stop unattended agent retry loops from hanging forever when a usage limit reset is too far away in Claude Code.
Add this to your agent environment or runner config:
1. Set CLAUDE_CODE_RETRY_WATCHDOG_MAX_USAGE_LIMIT_WAIT_MS to a number lower than your longest acceptable wait, for example 180000 for three minutes.
2. Keep CLAUDE_CODE_RETRY_WATCHDOG enabled so retries still run for transient errors and short rate limits.
3. When the limit reset time exceeds your threshold, the agent ends the request instead of sleeping past the session budget.
4. Log the termination reason in your orchestration layer so you can queue the task for the next scheduled run.
Bookmark this rule for your next agent session.
Grokverse retweeted
The Boring Company is building something far bigger than most people even realize
While everyone is focused on Tesla and SpaceX, Elon Musk is also transforming underground transportation and industrial infrastructure, and the scale of what’s now happening is incredible
The Boring Company is now valued at $23 billion after recently raising another $3 billion in funding
And that funding is helping accelerate something even massive: a new partnership to develop over 150 kilometers of underground infrastructure across the UAE, on top of the existing Dubai Loop project
Just look at what the company has already accomplished in Las Vegas:
• Over 4 MILLION passengers transported
• 14 operational stations
• 68 MILES of tunnels approved for expansion
• 123 stations approved across the planned network
• Targeting 28 operational stations by the end of 2026
And the expansion is already moving beyond Las Vegas
In Nashville, two Prufrock tunnel-boring machines are operating simultaneously, a third is scheduled to begin tunneling later this month, and a fourth is already being built
More than 40 stations are in various stages of planning and design
The company has also signed a new agreement to explore transportation and utility tunnels in Abu Dhabi
And just days ago, its massive 900,000-pound Prufrock-5 machine arrived in San Antonio to prepare for a new underground transportation project designed to improve accessibility
But here's another incredible example of how this technology is already being used inside Tesla's Gigafactory Texas
The Boring Company built an underground tunnel beneath an eight-lane highway to transport finished Cybertrucks directly from the production line to their staging area
Previously, that journey required around 12 minutes using surface roads
Now it takes just 60 SECONDS
That's a 12× improvement in transportation time
And the engineering behind the tunnels is even more impressive
The company has already demonstrated autonomous installation of massive concrete tunnel segments with millimeter precision, while its machines are designed to mine and install tunnel lining continuously
Think about what happens when this technology scales across factories, airports, cities and industrial infrastructure
And this is where the vision gets much bigger
The Boring Company is literally laying the foundation for 3D infrastructure for entire cities, where transportation networks expand across multiple underground levels instead of being limited to surface roads
Las Vegas, Nashville and Dubai are just the beginning
Elon Musk is fundamentally rethinking how cities will be built and how transportation will work at massive scale
And what he's building underground is going to be something far bigger than most people have ever imagined
Grokverse retweeted
NVIDIA expands the Nemotron family with a new flagship and two edge models, bringing open-weight agentic AI from data center to desktop.
Nemotron 3.5 Super VL is a 120-billion-parameter model optimized for specialized always-on agents, claiming fastest task completion and highest accuracy. It joins the open-weight lineup alongside two Bolt variants for local deployment: the 74B-A9B 24GB model balances intelligence and speed per GB, while the E4B 4GB model targets ultra-light edge scenarios.
The Bolt models are notable for fitting into consumer-grade hardware. A 24GB VRAM limit means many workstation GPUs can run the 74B variant locally. The 4GB E4B model opens possibilities for on-device agents without cloud dependency.
Open weights give developers full control over fine-tuning and deployment. NVIDIA is covering the full spectrum from server-scale agents to embedded inference, all under the same Nemotron umbrella.
Grokverse retweeted
Holy smokes... Claude Code v2.1.301 is out with a handful of quietly powerful changes.
• CLAUDE_CODE_RETRY_WATCHDOG_MAX_USAGE_LIMIT_WAIT_MS – unattended retry mode now kills a request cleanly when the next usage limit reset is too far out, instead of idling forever
• New % motion in vim mode for jumping to matching brackets, composable with d, c, y
• Mods get a rows field on Input: a multi-line terminal text area where Enter submits, Shift+Enter or backslash before Enter adds a new line
• OpenTelemetry tool_decision events now carry classifier_verdict and permission_mode to separate auto mode decisions from permission rules, plus classifier_category naming the matched built-in rule
• OpenTelemetry tracing setup guidance added to the claude-api skill, configurable with the ant CLI
For anyone running unattended agent loops, the retry watchdog fix alone is worth the upgrade – no more stuck processes waiting on a rate limit that won't clear for hours. Vim power users get the classic bracket jump, and mod authors finally have a proper multi-line input widget. The telemetry additions give teams auditable traces of exactly why a tool was blocked or allowed.
The quiet trend here: Claude Code is becoming a serious platform for building autonomous workflows, not just an interactive assistant.
Grokverse retweeted
Tesla didn't just show a new Roadster tonight. It showed a car that leaves the ground.
Elon said it best: "Turns out the future does have flying cars."
The top version is called Rocket. One seat, invite only, and it flies on SpaceX-style cold-gas thrusters. Tesla says it does 0-60 in under 0.59 seconds, goes past 230 mph and covers over 300 miles.
But the part that broke my brain isn't the flying.
It's the suction. A fan system pulls air out from under the car so hard that Tesla says it can hold the Roadster upside down while it's standing still. Grip before you even move.
And you don't need an invite to get in on it. There are three versions: Performance for the road at under 1.6 seconds, Thrust for the track at under 0.9 seconds, and Rocket for the sky.
Reservations are already open. Prices? Tesla hasn't said yet.
My take: every carmaker is chasing a faster 0-60. Tesla just changed the question. When a road car borrows hardware from a rocket company, the spec sheet stops being about horsepower and starts being about physics. Nobody else can copy this next year.
