🚀 Codex CLI 0.136.0 is out! 🔗 TUI markdown links stay clickable; wide tables reflow to key/value 📦 Archive/restore sessions from TUI or CLI 🪟 Windows alpha: codex sandbox setup --elevated Changelog:

🚨🇺🇸🇮🇷 BREAKING: The 60-day US-Iran deadline just ran out with no deal. The Islamabad MoU gave both sides until today to lock in a final agreement. Nothing got signed. They can't even agree on what the expiry means. Washington calls it the end of a ceasefire. Tehran says the MoU ended the war, so there was never a truce to extend. Hormuz is the real fight. Iran reclosed it, the US reinstated its blockade, and Trump's now floating declaring the strait U.S. territory. Everything points to more escalation. But six months in, both sides are exhausted enough that neither might have the appetite. Source: Al Jazeera / Writer: Daniyal

libghostty for WebAssembly got a lot of love recently and is now very, very, very good. I spent the day building harnesses and comparing to xterm.js and here are the results: faster at IO, faster at reflow, faster at rendering. Like, a lot a lot faster. We now provide pre-built `ghostty-vt.wasm` files in our GitHub releases. These are the fastest, most optimized builds that are compatible with every major browser going back a number of years. Before going over setups, I want to say that xterm.js is very good software. It is a relatively modern terminal, the maintainers are very nice, it's done incredible work for the web terminal ecosystem. I have nothing against them, but I'm in the business of making super fucking fast terminal emulators. Let me explain the setup in every case: IO throughput: This is an apples-to-apples comparison of `ghostty_terminal_vt_write` vs xterm's `Terminal.write(Uint8Array)` timed from first write to final completion callback. This is testing how fast a terminal emulator can process input and update internal state. Nothing else. Units are in MB/s processed. Note for graphemes: Ghostty has grapheme processing enabled and xterm.js does not (requires an addon). So, this shows how Ghostty does better despite having that cost. Column reflow throughput: Compares `ghostty_terminal_resize` to xterm's `Terminal.resize` with identical terminal states (same content, same size, same scrollback). The "dance" is rapid bigger/smaller resize of matching column widths, forcing text reflow. Units are in resizes/second. Render speeds: These test use the identical xterm.js WebGL renderer for both Ghostty and xterm.js. Since the renderer is identical, we can focus on how long it takes to prepare the inputs to the renderer (build the render state) and assert we get the same output byte-for-byte (since its the same renderer!). This is the truest way to measure these two because libghostty-vt itself does not provide a renderer, its about how long it takes to go from terminal state to render-ready. Full render: This tests a fresh renderer with nothing previously rendered. Units are in updates/second. Incremental render: This tests having prior state and only rendering specific changes such as adding scrollback, moving the viewport, moving the cursor. Units are in updates/second. This is using the latest released version of xterm in every case compared to the HEAD libghostty (since we haven't tagged a release, exact commit d760ee96e54657416eb427b793c7e839f003df7d).

Sharing some more UI previews for Superlogical, this time the "tab peek." If you three-finger swipe down, you can "peek" at the other session tabs. These are live-updating, metal-rendered previews including splits. If you keep dragging, you enter a "mission control" style mode. This is one of those UI patterns that once you get used to, you mindlessly find yourself doing it ALL THE TIME. It's a great way to quickly take a look at everything thats going on. You aren't reading the mini text of course, but the shape of it tells you a lot about where things are. The peek pushes the terminal down off screen, it doesn't resize it, so it doesn't force any weird reflow or anything either. There's a lot more to talk about, especially about how this interacts with how we expose what needs attention so you can easily find the terminals/sessions/etc. that require your attention. Of course, you'll be able to drive this with the keyboard too. :) As always, a disclaimer: this is all early work subject to change, just trying to share some of what we're building in public. A lot of the UI experiential design is @almonk so follow him for more updates too!

Another *giant* Grok Build release from @SpaceXAI 🔥 Version 0.2.121 (2026-08-05) is live. Here are the standout changes: Features - Dashboard rows now display a short summary of what the agent did on the previous turn - Extensions modal groups items alphabetically with collapsible Skills sections - Parent agent gets reminded to keep working when a subagent runs in the background - Clients can reattach to a running session without replaying the full transcript and can close sessions explicitly - Grok no longer asks which project directory to use when launched from home or other non-project folders - /feedback opens a dedicated report box instead of dropping into prompt mode - Auto theme detection works over SSH and inside tmux - Voice and Finance tool cards render with icons and localized labels - Markdown tables reflow inside cells on narrow panes instead of clipping Bug fixes include better handling of large MCP image outputs, reliable session resumes after remote restores, correct default branch detection, and smoother cancel behavior so background tasks stop cleanly.

🚨 NEW GROK BUILD UPDATE 🚨 v0.2.121 — 2026-08-05 Features: • Dashboard rows now show a short summary of what the agent did in the previous turn. • The Extensions modal now shows items grouped and sorted alphabetically with collapsible sections for Skills. • When a subagent runs in the background the parent agent is now reminded to keep working on the original task. • Clients can now reattach to a running session without replaying the transcript and explicitly close sessions. • Grok no longer asks which project directory to use when launched from your home folder or other non-project directories. • /feedback now opens a dedicated report box instead of entering prompt mode. • Auto theme detection now works over SSH and inside tmux sessions. • Voice and Finance tool cards are now rendered with icons and localized labels. • Markdown tables now reflow inside cells on narrow panes instead of clipping the right border. • Permission prompts now show the complete script clearly instead of dimmed fragments. • Long bash commands in permission prompts can now be expanded with Ctrl-F. Bug Fixes: • MCP tools that return images no longer drop or corrupt screenshots when output is large. • Resuming a restored child session after a remote parent restore no longer fails with 404. • The default branch is now correctly detected for hand-initialized repos that lack origin/HEAD. • Disabled MCP servers that can still be re-enabled now remain visible in the list instead of disappearing. • Rapid send-now presses or messages sent while waiting on subagents no longer lose earlier queued messages. • Pressing Esc or the stop button now reliably prevents background tasks from restarting the model after a cancel. • Login no longer skips when an invalid first-party API key is present in the environment. • Model picker and command palette now work while reviewing a plan before approving. • Workflow parallel() panels no longer launch hundreds of children at once and stall. • Dashboard overlay no longer shows useless prev/next shortcuts when only one agent is present. • Pinned prompt headers can now be selected and copied with the mouse. • Tab and Esc now behave consistently on every blocking card (question, permission, cancel-turn). • Dashboard navigation after /new now correctly exits back to the dashboard from an empty prompt. • Codebase restore no longer hangs on large or shallow git repositories. • Remote session resume restores conversation only unless --restore-code is explicitly used. • Copying CJK text with the mouse now includes every character at the selection edges. • Resume search now finds sessions by UUID even when they live in other directories. • API errors now appear as clean banners instead of raw JSON dumps in the TUI. • Typing exit or quit in the dashboard now exits the CLI instead of starting a new session. • Mode indicator (plan/agent/ask) now correctly reflects the session's actual mode after resume and transitions. • /delete now returns to the dashboard when you delete a session you opened from it. • Enter in the slash command menu now runs the highlighted command. • Grok now retries more server errors during outages for better reliability. • Syntax highlighting in long diff lines now stays correct when wrapped. • Slash commands that require a session now show a helpful message when used from the dashboard. • Exiting the CLI now properly resets terminal modes even in minimal mode. • Queued prompts now stay visible and reachable while waiting on subagents. • Auto recaps no longer appear in the middle of new turns or while busy. • Error messages in /btw side questions now show fully wrapped. • Queued slash commands and images can now be reordered in the queue pane. • Error messages in /btw side questions now show fully wrapped. • /feedback no longer resets the composer input mode after the pane closes.

My dear front-end developers (and anyone who’s interested in the future of interfaces): I have crawled through depths of hell to bring you, for the foreseeable years, one of the more important foundational pieces of UI engineering (if not in implementation then certainly at least in concept): Fast, accurate and comprehensive userland text measurement algorithm in pure TypeScript, usable for laying out entire web pages without CSS, bypassing DOM measurements and reflow

Grok Bot summary of SpaceX CFO Bret Johnsen’s new interview with Tony Kim at SpaceX Mission Control. Bret’s path - Came from semiconductors (Qualcomm, Broadcom; was CFO of MindSpeed) - Barely knew SpaceX / Elon when a board member called - Thought aerospace sounded “old and stodgy” vs 6-month chip cycles - Visited the rocket factory, called his wife on the drive home: we’re moving back to LA - Now sits in the same magical factory: Dragon in the rafters, ~700 launches run from Mission Control, ~40 buildings feeding the campus - Full circle: semiconductors again via AI / chips --- Taking on AI complexity (during the biggest IPO ever) - SpaceX DNA: run toward “impossible” challenges (first booster recovery felt like a rock concert) - xAI acquisition in January + X + Cursor + Colossus = complexity multi-X in ~1–2 years - AI is now the biggest business at SpaceX - Bought time-to-market into the biggest product cycle of their lifetime - SpaceX leadership DNA applied to that team - Already feel leading on infra cost and speed - Products moving: Grok 4.6, 4.7 soon, Grok Bot viral, Cursor strong - Every week matters; soon every day --- AI business: the money picture - ~$100B company ARR exiting this year (annualized on December) — AI is more than half / the majority - A year ago AI revenue was ~zero - Capacity: a bit over ~2 GW this year → guidance 5–10 GW by end of next year - Becoming one of the biggest hyperscalers, fast Why they move so fast (vertical integration) - Engineer the data centers themselves - Act as GC: buildings, power, cooling, networking (plus partners) - Used to regulators from rockets + telecom - Removes margin stacking - Talent transfer: launch-site / pressurized systems people onto data centers — Elon compared it to Yankees vs little league Hosting model / ROI - Monetize compute while catching up on software - Demand knocked: “what are you doing with all that Colossus compute?” - Short contracts: ~3 months + 3-month out (flexibility for their own future training needs) - Payback: dramatically less than one year on cutting-edge terrestrial compute - Hosting customers include tier ones like Anthropic, Google, and others - Same watts can later flip to their own software at higher $/watt — optionality Frontier models - Moving maybe faster than anyone on trajectory - ~5 serious players; need to be at frontier for recursive improvement - Coding solutions improving dramatically - Consumer vs enterprise tools will stratify - Grok Bot: growing faster than anything they’ve seen; “transformational” user feedback; enterprise coming; expects roll into Cursor AI safety - Follow Elon’s self-regulation / peer-review path (like other industries) - Logical, collaborative, including outside the US - Hopeful something lands quickly TAM framing (IPO) - ~$28T TAM quoted at IPO; AI was ~$25–26T of that (digital AI only) - Physical AI not even in that number — Bret thinks physical AI eventually bigger than digital - Launch / space TAM = hundreds of billions (almost rounding vs AI) - Connectivity (mobile + broadband) ~$2T — still small vs AI - Differentiation: infra + X content (~500–600M MAUs) + consumer + enterprise push (Cursor etc.) What to watch: Starship + terrestrial compute Bret’s prioritization cheat code: 1. Get Starship right 2. Get terrestrial compute right for the next ~2 years until orbital kicks in → Transformational financially Starship status - Catalyst for all three businesses: space, satellites, AI - Holy grail: full reusability → another ~10x cheaper $/kg to orbit - Opens moon boots / lunar base / bigger industry - Flight 14 later this month: production V3 Starlink broadband satellites — real revenue payload - One Starship V3 load ≈ ~20× what Falcon 9 flew for Starlink capability - Each V3 satellite ~1 terabit; ~50–60 per flight = enormous bandwidth per launch - A couple launches start filling V3 planes; first-half-next-year benefits - Eventually Starlink as % of global internet traffic - Connectivity demand from machines (robots, AVs, agents) will dwarf human brains - Same Starship unlocks next-gen direct-to-cell / direct-to-device - Same bus + more solar + radiator + compute chips = orbital compute constellation Reusability timeline - Flight 15 may attempt catch / bring back first + second stage - First stage already recovered and reflown - Flight 13 soft splash best yet; recovered ship like a whale, engineers studied heat shield - Reliable stage recovery this year; meaningful reflight next year - Once a few second stages reflight → orbital compute cost can match terrestrial - Then orbital cost curve goes down (tech / Moore’s law / aircraft-like amortization) while terrestrial gets more expensive (power, labor, regulation, cooling) - Clean free solar, no community impact - Industry agrees space compute is the future — debate is timing; others say ~decade; SpaceX flies first ones next year; meaningful compute in ~2028 Scale ambition - Both cost and speed matter; near-term availability/speed most important - On track toward ~100 GW/year by end of decade (~20% of entire US grid each year) - Elon talking terawatt-scale in the 2030s - Hosting proves staying power + virtuous cycle for industry - New NVIDIA platforms: just fly the next satellite generation — no whole new terrestrial campus cycle TerraFab (chips / memory) - Parallel to orbital ramp, not after - Not replacing NVIDIA — love the partnership; will buy all they can - Worry is deeper fab / memory supply chain - Memory industry ~30%/yr growth vs SpaceX wanting triple-digit — won’t work - Hard technical challenge; memory may be early focus - Want volume benefit by late decade (demo next year with Tesla development line → something in ’28 → meaningful ’29+) Viewer Q&A highlights **First bottleneck in the AI chain?** - Chip supply for next year - Power for infra players (utilities used to ~1% growth; turbines / blades constrained) - SpaceX standing up investment casting for turbine blades themselves - Orbital is the big unlock; squeeze every terrestrial watt meanwhile How to tell durable AI spend vs keep-up spend? - Bret: durability of revenue; diversified monetization of the same compute; host now, software later at higher $/watt; guided next-year monetization range ~$30–50/watt - Tony: stack layers have different costs, margins, contract duration; SpaceX spans many layers with <1-year payback on infra How soon a data center in space? - First orbital compute satellites end of next year - Meaningful compute in 2028 - Better phrase: “racks in space,” not giant buildings - Same V3-like bus + solar + radiator + compute chips - Already flown ~11,000 sats; launch + inter-sat + ground links already solved - ~full rack (~72 NVIDIA-class) concept per sat; ~5 MW compute per launch Tony’s investor wrap - Still early; trends accelerate into physical AI and beyond Earth - Connectivity / tokens increasingly for machines, not just humans --- Bottom line in one breath SpaceX is using rocket DNA — vertical integration, crazy talent, short-cycle iteration — to become a top AI infra + software player now, while Starship unlocks Starlink V3, direct-to-cell, and orbital “racks in space.” Watch two things: terrestrial compute scale (toward $100B ARR and multi-GW) and Starship reusability. Everything else is downstream of those.

The 14th flight of Starship is preparing to launch as soon as Tuesday, September 22, pending regulatory approval. The 75-minute launch window will open at 7:15 a.m. CT. According to @SpaceX “The upcoming flight is planned to be the first to send Starship into orbit around Earth. Flight tests until this point have intentionally flown passively safe suborbital trajectories to maximize public safety while allowing for maximum learning.” “This will also mark the first time we plan to deploy Starlink V3 satellites into the constellation, delivering a payload that will dramatically expand connectivity speeds and reliability around the world.” Starship’s initial orbital mission is expected to fly at an altitude approximately 275 km above Earth and complete approximately six orbits around the planet over the course of a nearly 10-hour flight, with splashdown targeted in the Pacific Ocean to the west of Chile.” “The booster’s primary test objective on Flight 14 will be executing a successful launch, ascent, stage separation, boostback burn, and landing burn at an offshore landing point in the Gulf of America. There have been several modifications to hardware and software to address issues seen on the previous flight.” “After stage separation and flip on Flight 13, the Super Heavy booster was able to use all 33 engines on its boostback burn for the first time. In the terminal phase of the burn, the three center engines showed signs of ice clogging which triggered an early end to the maneuver. The booster went on to attempt a landing burn, with 8 of the 13 planned engines reigniting before the booster made a hard splashdown in the Gulf.” “The Super Heavy on this upcoming flight has hardware modifications to improve filtering to the engines and software changes to enhance relight reliability.” The Starship upper stage’s primary objectives include the first orbital insertion maneuver, the deployment of 26 Starlink V3 satellites, a deorbit burn using a single Raptor engine while in space, and a controlled reentry, descent, and splashdown in the Pacific Ocean.” “Starship will only execute a burn to enter orbit after the flight control team has ensured there is sufficient redundancy on hardware critical to doing the subsequent deorbit burn at the end of the mission. Starship will deliver 26 Starlink V3 satellites to orbit for the first time, which aim to greatly expand the network's capacity and user speeds. 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. “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.” “Several upgrades and experiments related to Starship’s heatshield will also be tested, with some improvements derived directly from data gathered from the Flight 13 Starship as it floated in the Indian Ocean.” “They include additional retention mechanisms added to tiles in areas deemed to be at highest risk of falling off during ascent, addressing recently discovered areas that offer flow paths behind tiles for plasma, and flying multiple areas with a curved tile design that has shown the ability to reduce heating in the gaps between tiles.” “And finally, two tiles recovered from Ship 40 are planned to be reflown on Ship 41, marking the first tile reuse for Starship.” @esherifftv