Grok Build
Grok Build just got a pretty substantial upgrade across MCP, memory, long-running sessions, and overall agent reliability
MCP tools can now return structured JSON, cancelled tool calls actually tell the server to stop working, memory can be managed directly from the /memory browser, and long-running sessions keep their compaction checkpoints instead of losing them during cleanup
There are also fixes for macOS image pasting, subagent cancellation states, /rewind, huge skill files flooding context, Windows cloning, and minimal-mode session handling
A lot of edge cases cleaned up here...especially for people running heavy agent workflows for long periods
Release Notes: v1.0.33
Features:
• MCP tool results now include structured JSON data when the server provides it.
• GROK_GROVE=1 now enables both clone and worktree features when specific knobs are unset; missing remote settings no longer force copy.
• Delete memories from the /memory modal using two-press x; only actual notes (not generated indexes) can be removed.
• Composer footer now shows the Alt+Enter chord for inserting a newline when Cmd+Enter is unavailable over SSH.
• /memory now always opens the memory browser modal; status and toggle are available via keys inside the modal instead of command arguments.
Bug Fixes:
• Fixed a bug where pasting an image on macOS could attach the wrong image.
• MCP tool calls listed directly now show their name, arguments, and any error message.
• Background subagent tasks now correctly show cancelled status when the parent session closes.
• The Subagents dock no longer shows an unclickable "Done N completed" row after all subagents finish.
• Cancelled or timed-out MCP tool calls now notify the server so it can stop work instead of continuing in the background.
• /rewind no longer fails when older compaction checkpoints are missing after a session sweep; only the base checkpoint for the target matters.
• Large skill files are now capped at the same 25k-token limit as other reads; oversized skill bodies show a truncation note instead of flooding context.
• Long-running sessions no longer lose old compaction checkpoints or prompt offloads during the 30-day cleanup sweep.
• Auto "where was I" recaps no longer appear while a scheduled task, monitor, or workflow is still running.
• In minimal mode, confirming /delete or /exit now immediately opens a new empty session instead of appearing stuck.
• In minimal mode, resuming a session after /new now correctly reloads its history instead of showing the wrong session.
• MCP tools that take no arguments no longer show a stray "{}" line in the permission sheet.
• Fixed Windows clone and attach failures when .git/config values contain backslashes, quotes, or other characters that must be escaped.
• /memory modal now shows clear explanations when memory is disabled and helpful empty states instead of blank panes.
• /memory (and on/off toggles) no longer leaves user bubbles or host turns in scrollback when opening or changing the modal.
• Fixed Windows clone staging directories that could not be cleaned up after deadline or cancel during shallow bootstrap.
609
898
5,034
3,092,871
Elon just gave one of the clearest explanations yet for why Terafab has to exist
There are basically two problems:
• Geopolitical risk — at some point, advanced chips from Taiwan may not be available
• Scale — even if supply remains stable, existing fabs eventually run out of capacity as AI, humanoid robots and autonomous cars keep scaling
And chips are the foundation underneath all of it
Server AI, edge compute, Optimus, Tesla vehicles....everything needs enormous and reliable semiconductor supply
Elon:
“So it’s either build Terafab or fail to scale”
That pretty much sums up the entire strategy
Secure the chip supply....then scale AI and robotics without a ceiling
All-In Summit: Elon Musk & Gwynne Shotwell
-- AI Risks and Peer Review
-- Starship
-- Terafab as Taiwan Insurance
-- SpaceX/Tesla Merger Potential
-- Management at SpaceX
++ much more!
(0:00) SpaceX's @Gwynne_Shotwell joins The Besties!
(1:45) Gwynne's SpaceX story, selling rockets, and working for Elon
(6:45) Running SpaceX: AI, Starlink, Rockets, and X
(13:43) Direct to cell with Starlink, retiring rockets, competition
(18:34) Data centers in space
(22:00) Management at SpaceX
(27:37) @elonmusk joins: AI's real risk, model peer review, what he meant by "Dario is right"
(36:53) Elon and Gwynne on their working relationship, Starship's future
(46:16) Terafab, Tesla, Flying cars?, merging Tesla and SpaceX
(52:51) Lying AIs, how to do AI peer review right
$TSLA $SPCX
196
469
3,170
735,955
SpaceXAI engineer Lingxi Li, @lingxi, runs code cleanup while everyone else is asleep. Nobody is shipping at 3 AM, so conflicts are rare and the changes are low risk.
His bots start a research agent that sweeps the monorepo for quality issues, poor modularization, comment bloat and security gaps. He wakes up to a stack of PRs.
Three SpaceXAI employees are building a company in 3 days with Grok Bot. This is Day 1.
Matt Palmer (@mattyp), Lauren Tan (@poteto), and Roshan Sadanani (@roshan_s) start with research, a plan, and a product.
Live now, plus sessions for engineering, product, and founders.
nitter.click/i/broadcasts/1AxRnZbVp…
162
352
3,220
933,786
Elon Musk retweeted
Finally!!! A proper orbital attempt! Important note is that they won't be targeting orbit at SECO. They will only attempt an orbital circularization burn if all systems are go and it's considered safe. Nice and conservative approach!
Starship is preparing to go to orbit. Flight 14 is targeting to launch as early as Tuesday, September 22, pending regulatory approval → spacex.com/launches/starship…
177
438
6,744
894,388
Elon Musk retweeted
“Cisgender. It’s a way of marginalizing a normal person”
-Norm Macdonald
123
926
16,827
916,083
Elon Musk retweeted
Love seeing government officials actually trying @Tesla FSD (Supervised). Here's the Deputy Prime Minister of Slovenia using it on public roads:
"t doesn't get tired, doesn't fall asleep, and can significantly contribute to safer driving. On to the next step – robotaxis."
Jernej Vrtovec
117
521
4,982
501,504
Elon Musk retweeted
What’s crazy is that this bandwidth growth is well known but it’s likely impact is totally dismissed at the same time.
For context here is Starlinks historic bandwidth growth by generation. The V2 -> V3 inflection will 3-10x greater than the V1 -> V2 inflection was on this video.
Starlink V3 constellation starts being deployed this month.
Ultimately, it will deliver more than 100 times the bandwidth of our current 11,000 satellite constellation.
55
206
1,619
612,589
Elon Musk retweeted
Vice President @JDVance had a simple idea: Why don’t we have a Most Wanted Fraudsters list?
So we built one.
In just 3 months: 5 fraudsters captured, 4,000+ days on the run, nearly $2 BILLION in alleged fraud—across 3 continents.
The American people remain one of the FBI’s greatest partners.
You give us the information. We find them. We bring them to justice.
-DKP🇺🇸
2,730
9,179
49,521
997,453
If you're a fan of Mach33 research and you're not plugging the MCP/Connector into your agents (my preferred is grok bot) then you are falling behind.
just a few moments ago, 20 days after a very quiet release, total research interactions from MCP eclipsed total research visits on our site over the same time.
The median user on the site has read 2 analysis over the last 20 days and the median user on the MCP has made 36 analysis calls. 18x more depth of engagement.
the really nutty thing is that this trend is actually meaningfully accelerating over the same amount of time, across free users and paid users.
this does not include internal users which would dramatically accelerate this trend.
do yourself a favor and connect the Mach33 MCP to grok/claude/chat gpt whatever you use.
research.33fg.com/mcp-guide
You're agent will get much smarter on any space/spacex related question.
561
902
3,964
3,007,192
Elon Musk retweeted
I have used my X Money virtual Visa card, via my iPhone wallet, twice today. Works perfectly.
This app has come a long way from the days of Twitter.
184
301
2,098
374,660
Blizzard’s inability to solve the login problem on their product launch day reminds me of the classic Forster essay: The Machine Stops
cs.ucdavis.edu/~koehl/Teachi…
1,236
1,285
16,340
6,155,679
Three SpaceXAI employees are building a company in 3 days with Grok Bot. This is Day 1.
Matt Palmer (@mattyp), Lauren Tan (@poteto), and Roshan Sadanani (@roshan_s) start with research, a plan, and a product.
Live now, plus sessions for engineering, product, and founders.
nitter.click/i/broadcasts/1AxRnZbVp…
651
1,498
10,113
3,976,783
Starlink V3 constellation starts being deployed this month.
Ultimately, it will deliver more than 100 times the bandwidth of our current 11,000 satellite constellation.
Starship's upcoming 14th test flight will deliver 26 Starlink V3 satellites to orbit for the first time. Each Starlink V3 satellite will add 1 Tbps of capacity to the constellation, for a total of 26 Tbps of capacity added on this mission alone. That’s ~10x the capacity compared to a single launch of V2 mini Starlink satellites on Falcon 9.
After deploying from Starship, the Starlink V3 satellites will unfold their antennas and deploy their solar arrays and make initial contact with the ground and the rest of the Starlink constellation via radio frequency and laser links. The satellites will then begin raising their orbits with their onboard thrusters. Following their on-orbit checkouts, the satellites should begin serving customers as soon as a few weeks after launch!
Three of the satellites have been modified with a suite of cameras to scan Starship’s heatshield and transmit imagery down to operators to continue testing methods of analyzing Starship’s heatshield readiness for return to launch site on future missions.
992
2,558
20,240
3,908,682
