well. I just had a 36 minute call with my @bot resulting in him launching @cursor_ai cloud agents to build a workflow that tests automated phone support agents by having @grok voice pretend to be a customer. meanwhile cars are doing flips and my bed is notifying me that it is cooling itself down a little extra to keep me comfortable. it’s fun to be living in the future.

Important note regarding Grok @Bot: Going forward, @SpaceX will use the best back end model for any given task, including Claude Opus 5.5, MidJourney, Suno and other leading APIs. Whatever is most likely to give you the best outcome.

When a nerd tells me they don't use agents "because I can write better code than any agent", I usually go "huh, sure buddy, you do you". When they tell they don't use agents for monitoring, I just laugh in their face. Seriously. You can't monitor a production deploy better than an agent can. I dunno about AGI but this specific job is something that an LLM can already do better than every single human in the world. For me, as a systems engineer, having agents monitor my infra changes has been (no joke) more impactful in my day-to-day work than having agents write my code. I started doing this many months ago by asking the agent to monitor explicitly after I triggered a deploy; soon after I moved to using a set of my own Skills. Now our Review Agents team has fully automated this workflow with delightful UX and obviously I'm not going back. It's hard to overstate how powerful this is. All it takes, for me, in production: - The Datadog MCP - The PlanetScale MCP (for all our databases) - Our custom MCP for deploys and internal tooling That's it. Some carefully designed gradual deploy pipelines, and the agent just takes care of everything from there. You can't really deploy SEVs to production anymore. Doing infra like this is pure joy. And the safety is just a tiny part of it. The Rollout agent also does the actually hard part of my job, which is _performance_. When the changes you're deploying are a performance optimization, the agent prepares a monitoring plan that ensures the optimization is correct AND effective. I rarely prompt benchmarks for my optimizations anymore. Just fucking go to prod and the agent is gonna give me the best possible benchmark. If you work on systems or infra for a living, you should be setting up your production environments so you can use Rollouts. Not doing so is actually reckless.

All seven Databricks co-founders began our journey together at UC Berkeley. We started the company in a room in Soda Hall, moved to a small office on Addison Street, and spent our first few years in Berkeley before eventually heading to SF. Two of our co-founders are still on the faculty. Today we announced Databricks Field at California Memorial Stadium, our first collegiate athletics sponsorship. Berkeley shaped everything about how this company thinks, and this is our way of investing in the next generation of students and builders who'll do their best work on that campus. Go Bears!

here's a prompt to improve your agent harness based on what we've learned at cursor. enjoy # Improve this agent harness's token efficiency You're working on an LLM agent harness: the system prompt, tool definitions, request assembly, context caching, compaction, and retrieval, and how work is split across agents. Make the agent's runs cheaper without making it worse at its job. - Objective: lower price-weighted token cost per completed task. - Constraint: no measurable drop in task quality. Measure per task, not per request. Every turn resends the prefix (tools, instructions, setup, and the conversation so far), so a change that shrinks each request but adds turns can cost more. Weight tokens by billing type: output, uncached input, and cached input are priced very differently. Work in this order: map the harness and measure the baseline, rank the opportunities, make the changes that are safe to make directly, put the rest behind flags or in proposals, then report. Figures below come from one team's production coding agent and its multi-agent experiments. Use them to gauge magnitude, not as targets. One round of these changes (prompt trimming, tool offloading, cache layout, sparse line numbers, subagent tuning) cut that team's overall token cost about 7% with no loss in quality. The larger percentages apply only to the part of the request each change touched. ## Principles 1. Change what the harness sends, not how hard the model tries. Don't ask the model to conserve tokens. A harness that told its model to "take care to preserve tokens and not be wasteful" found it grew reluctant to take on ambitious tasks and sometimes quit, saying it wasn't supposed to waste tokens. 2. Capable models need definitions, not commands. Lists of "DO NOT", "You must", and "Important", and guards against older models' habits, can usually be replaced with plain descriptions of what each tool does. One team cut about two-thirds of its system prompt this way, and the shorter prompt worked across model families. Instruct only on what the model can't know (the product, the environment, the user's processes) and on quirks you've seen in transcripts. 3. Static context is for what most turns need. Everything else should be discoverable when needed. Less up-front context also means less confusing or contradictory information. 4. Expect removals to win. Guardrails written for weaker models, coordination steps that became bottlenecks, and prompting for behavior the model now does on its own all cost tokens. 5. Real usage decides. Evals are a fast proxy, but they skew toward hard problems and miss the real mix of requests. ## 1. Map the harness and measure the baseline Find: - Where requests are assembled, the system prompt, and tool schemas. If a framework or SDK builds requests, find its hooks for message order, cache control, and tool loading. - How tool results are formatted, and how history is kept, trimmed, or summarized. - How subagents or parallel agents are spawned, if any. - Which models and provider APIs are used. From the provider's docs, get the prompt caching behavior (automatic or explicit breakpoints, TTL, minimum cacheable length) and the prices for output, uncached input, and cached input. - Existing logging, token accounting, and evals. If the harness doesn't record per-request token usage by billing type and cache hits, add that first. Everything later depends on it. Then render a few real requests (from logs, or by running representative tasks) and count tokens per section with the model's tokenizer or the API's usage fields. Produce: - Cost share by source × billing type. Sources: system prompt, tool definitions, skill/rule/integration descriptions, user messages, file reads, search results, command and other tool output, history, summaries, subagents. - Static tokens per request, cache hit rate, and turns per task. - Per tool: the share of runs that call it at least once, and its error rate. Read the rendered requests, not just the templates. Duplication, leaked volatile values, and misordered blocks only show up there. Rank opportunities by share of spend × fraction removable ÷ quality risk. ## 2. System prompt and injected context Label every instruction: - Keep: product or environment knowledge the model can't infer, fixes for quirks seen in this model's transcripts, and rules a mode depends on. - Rewrite: commands and emphasis into plain descriptions. Reminders into constraints: "No TODOs, no partial implementations" works better than "remember to finish implementations." Vague quantities into ranges: "generate 20–100 tasks" gets far more ambitious behavior than "generate many tasks." - Delete: things capable models do by default, guards against behavior you haven't seen from this model, text that repeats tool descriptions, and lines that could contradict a user request. Models trained to rank system instructions above user messages will side with the system prompt. - Move: anything per-user or per-request (date, environment, repo state, lists of skills or subagents, user rules) into a user-role setup message after the cache boundary. Audit other injected context the same way. As models improved, the team behind these figures dropped directory trees, pre-retrieved snippets, compressed copies of attached files, lint errors injected after every edit, forced expansion of short file reads, and caps on tool calls per turn. They kept small, high-value facts: OS, repo status, and open or recently viewed files. Skip checklists for open-ended work. The model optimizes the listed items and deprioritizes everything else. ## 3. Tool definitions Tool schemas ride along on every request. Most tools beyond the core set were each needed in under 20% of conversations, and moving them out of static context cut tool-description tokens 60%. Doing the same for integration tools (such as MCP servers), with names in context and full schemas in one folder per server that the agent can search with grep or jq, cut total tokens 46.9% in sessions that used them. - Keep in static context: high-frequency tools (for a coding agent: read, search, edit, shell), tools the model tries to call even when they're absent, and tools a mode depends on. - Offload the rest: leave a name or one-line pointer and make the full schema discoverable on demand. Group related tools so they load together, and put status (such as "needs re-authentication") where the agent will see it. - Tighten what remains: describe behavior and arguments, and drop usage lectures. - Pick the split by testing a few configurations and tracking tokens, cost, latency, tool-call errors, and task success. ## 4. Cache layout Order each request so the reusable prefix is as long as possible: `tool definitions → system instructions → [breakpoint] → setup message (skills, subagents, rules, environment) → [breakpoint] → conversation` - Keep the prefix byte-identical across turns. Use deterministic tool order and serialization, put timestamps and IDs after the boundary, and don't rewrite earlier messages except when compacting. - Use explicit breakpoints if the provider supports them. Otherwise rely on automatic prefix caching with the stable part first. Respect TTL and minimum-length rules. - Switching models mid-conversation throws away the cache (caches are per model and provider) and hands the new model a history it didn't write. When a different model is needed, run it as a subagent with fresh context. Explicit breakpoints plus moving per-request setup after them cut cold cache misses 20%. ## 5. Tool results and other context added during a run - Large outputs (commands, integrations, logs): write them to a file and return the path, size, and a short tail. The agent can tail, grep, or read ranges for more. Truncating loses data, and inlining bloats every later request. Treat long-running terminal sessions the same way. - High-volume formats: look for overhead repeated on every line or item. Numbering every 10th line of a file read instead of every line cut cache-read tokens 1.6% without hurting citation accuracy. Each number costs 3–5 tokens, and agents read tens of thousands of lines per session. Also check repeated absolute paths, verbose JSON keys, ANSI codes, progress bars, and repeated headers. - Good retrieval saves exploration turns. Adding semantic search alongside grep raised codebase question-answering accuracy 12.5% on average and cut the iterations users needed. - Tool errors waste tokens and leave confusing debris in context. Classify expected errors (invalid arguments, unexpected environment, provider error, timeout, user abort), treat unknown errors as harness bugs, and track rates per tool and per model. One focused effort along these lines cut unexpected tool errors 10×. ## 6. Long runs: compaction, subagents, and model mix - Compaction: keep the summarization prompt short and the summary compact, carry forward plan state and remaining tasks, and save the full history to a file the agent can search for details the summary dropped. A model trained to self-summarize from a one-line prompt wrote ~1k-token summaries with half the compaction error of a multi-thousand-token prompt that produced 5k+ token summaries. Untrained models may need more guidance, so test how short you can go. A more expensive summarization model made a negligible difference. - Scratchpads and running notes: rewrite them instead of appending. For repeated work in one environment, a small agent-maintained notes file with a line budget, loaded at start, is a promising way to shorten later runs. - Subagents: fresh context keeps the parent lean, but isolation adds coordination cost (duplicate or stale work). If the model already delegates on its own, remove prompting that pushes it to. Have subagents return short handoffs: what was done, findings, concerns, and deviations. A subagent should use a different model only when the user or harness says so. - Model mix: in large multi-agent runs, workers used at least 69% of tokens, and over 90% in most runs. A frontier planner with cheap workers matched a frontier model doing everything at about one-eighth the cost. Planner choice still changes worker spend. One planner that cost less on its own saw its workers use several times more tokens, and the run cost more overall. Measure the whole tree. - Routing and reasoning effort: send simple turns to a cheaper model or lower effort, and upgrade only when a stronger model is clearly better. A router built this way matched or beat single frontier models on user satisfaction at 41–68% lower cost. - Reasoning continuity: if the API returns reasoning items (including encrypted ones), pass them back on later turns and alert when they go missing. Dropping them cost one reasoning model 30% on a coding benchmark, and it burned tokens reconstructing its plan. ## 7. Fit the harness to each model Adapt to what each model was trained on instead of forcing one shape on all of them. If you've tuned the harness for a similar model, start from that version. - Edit format: use the one the model was trained on (for example, patch-style or search-and-replace). An unfamiliar format costs extra reasoning tokens and causes more mistakes. - Shell or tools: shell-first models fall back to `cat` or inline scripts. Name tools after their shell equivalents (such as `rg`), and if needed add: "If a tool exists for an action, prefer to use the tool instead of shell commands (e.g. read_file over `cat`)." - Literalness: some model families follow instructions literally and others tolerate imprecision. Some spiral on emphasized wording. Strip caps and emphasis for literal models. - Triggers: some models ignore a tool until told when to use it. A literal trigger works: "After substantive edits, use the to check recently edited files for linter errors. If you've introduced any, fix them if you can easily figure out how." - Progress updates: if a model reports progress through reasoning summaries, keep them to 1–2 sentences that note new findings or a change of tactic, and remove instructions about messaging mid-turn. - Quirks worth a targeted line: hedging or refusing as context fills ("context anxiety"), declaring completion early, stopping to ask permission, and calling tools that don't exist. Tie each added instruction to the transcript behavior it fixes. Re-audit when models change, since guidance one version needed can be dead weight for the next. ## 8. Validate - Offline: run a fixed set of realistic tasks before and after, ideally drawn from real usage and phrased the way users actually write (short and ambiguous). Compare task success, tokens, cost per task, turns, and tool errors. Don't ship a change that lowers success. - Online, if you have users: A/B test each change or small bundle. The primary metric is cost per completed task. Guardrails are task success signals, tool-call errors, latency, turns per task, and cache hit rate. For a coding agent, a good success signal is how much agent-written code survives over time. In general, check whether the user's next message moves on or reports a problem. - Ship only when cost drops and no guardrail regresses beyond noise. Record null results. ## What to change directly and what to propose - Change directly, each in its own revertible commit: token and cache telemetry, deterministic serialization and tool order, moving volatile content out of the cached prefix, explicit cache breakpoints, writing large outputs to files instead of truncating, passing back reasoning items that are being dropped, and fixes for recurring tool errors. - Change behind a flag so it can be tested: system prompt edits, tool offloading, output format changes, compaction changes, and subagent prompting. - Propose only: changes to which models run, routing, reasoning-effort defaults, or how work is split across agents. ## Traps - Asking the model to use fewer tokens or do less. - Truncating tool output. - Dropping reasoning items to save input tokens. - Volatile content in the cached prefix, or tool order that changes between requests. - Offloading a tool the model needs on the first turn or tries to call when it's missing. - Emphasis-heavy prompts (MUST, NEVER, IMPORTANT, all caps), especially with literal models. - Forcing a terser output format than the model was trained on. Fewer output tokens can mean less thinking and worse results. - Optimizing raw token counts instead of cost, per request instead of per task, or evals instead of real usage. - Switching models mid-conversation to save money. - Adding coordination layers that become bottlenecks. ## Report back with 1. The harness map and baseline: cost by source × billing type, with the biggest sources called out. 2. A ranked list of changes: layer, what changes, estimated savings and how you estimated them, quality risk, how to validate, and how to roll back. 3. The changes you made, including a system prompt diff with a keep, rewrite, delete, or move reason for each line. 4. A test plan for the flagged changes. 5. Gaps: anything you couldn't find or measure.

1 step closer to fully automating my job! I just created a @bot in my main @spacexai account that set up the Grok Bot app on its own machine and is now fully managing my demo account. All from my phone on a flight with shit wifi.

I don’t think people are thinking nearly big enough about what AI could actually mean for science. Most of the conversation is still about jobs, chatbots, writing, coding, images, productivity etc. But some of the biggest changes may come from AI helping us solve problems humans have been idle on for decades or even centuries and simply haven’t been able to crack. Think about how much of modern life ultimately traces back to a relatively small number of humans figuring out something fundamental about reality. Newton understanding motion and gravity. Einstein changing our understanding of space and time. People figuring out electricity, evolution, germ theory, DNA, computation. A human notices something. Spends years and years thinking about it. Connects ideas nobody had connected before and suddenly entire branches of civilization become possible. Like the aqueducts in Rome… Now imagine having systems that can read basically the entire scientific record, analyze enormous datasets, run simulations, compare millions of possibilities, notice patterns humans missed and do it across EVERY field at once. There are mathematical problems people have spent their entire lives trying to solve. There are parts of the ocean we barely understand and species we haven’t discovered. Huge portions of the rainforest we haven’t properly catalogued. Biological systems doing absolutely insane things that we still don’t barely/fully understand… For example, there’s a jellyfish that can essentially revert back to an earlier stage of its life cycle instead of continuing toward death. Animals that regenerate limbs and organs. Creatures that survive radiation, freezing, crushing pressure or almost no oxygen. Nature has already solved an unbelievable number of problems that humans haven’t. We just don’t necessarily understand how to translate those solutions into something useful for us. What happens when AI can map millions of species, proteins, genes and biological mechanisms and start finding connections between them that no individual scientist could ever hold in their head? Maybe something an animal evolved 50 million years ago eventually teaches us how to regenerate damaged human tissue. Maybe some obscure organism in the rainforest contains a mechanism that becomes a cancer treatment. Maybe studying aging across thousands of species reveals something we’ve completely missed about why our own bodies age. I’m thinking Jurassic park was a documentary on the future. Then there’s physics. Dark matter. Dark energy. Quantum gravity. What actually happens inside black holes. Whether there are forms of matter or physical phenomena we haven’t discovered yet. Whether our current understanding of space and time is even close to complete. Maybe time travel as we imagine it is impossible. Maybe it isn’t. The point is that we genuinely do not know what a more complete understanding of physics eventually makes possible. Imagine explaining GPS, nuclear power, an iPhone or landing something on Mars to someone living before Newton. It would sound like magic. That’s the part I keep thinking about. We are extremely used to living inside the discoveries humanity has already made, so we forget there could still be enormous layers of reality we simply haven’t figured out yet. AI doesn’t magically know the answers. We still need experiments, evidence, scientists and the physical world to tell us when an idea is actually true. But for the first time we may be building a tool capable of searching through the space of possible ideas at a scale no human mind ever could…and if even ONE of those discoveries ends up being on the level of electricity, antibiotics, relativity, or computing, the world on the other side of it could be almost impossible for us to imagine from where we are now.