The Computomatix Times

Training tiny models for special purpose use cases works so incredibly well if you have a great self improving recursive flywheel. Shopify ML team is on fire.

finetuned 0.8b model beats GPT 5.6-sol xhigh in this very specialized task.

I'd always thought AI was terrible at design, but after reading today's 🤯 post by @anshuc, I realized I was just doing it wrong.

"AI models are capable of amazing creativity, but that creativity gets stifled. LLMs are trained to be next-token predictors: they look at a sequence of text and predict what typically comes next. Great design is exactly the opposite of this. Great design bends the rules and delights users with memorable, unexpected choices."

@anshuc led design and engineering teams at Apple for 12 years. In his words: "Most people only see 1% of AI's creative potential. I want to show you how to tap into the other 99%."

His 8 techniques for breaking out of the 1%:
1. Use seed strings to inject variety
2. Be much more ambitious with your prompts
3. Create positive feedback loops with subagents
4. Use image generation to enrich designs
5. Use video generation
6. Cut out elements that don’t add value
7. Remove AI tells
8. Rewrite copy by hand

Read the post here: lennysnewsletter.com/p/how-to-turn-your-ai-int…

P.S. This design was made by AI 👇

Today is a very historical moment for AI video generation

You can now generate AI video faster than you can watch it

Before it'd take let's say 2-5 minutes to generate 15 seconds of video

@fal made a post-trained Minimax H3 variant called Max which is 50x faster than the original but still maintains quality

It generates 15 seconds of video in 9 seconds!

That means you can now do new things like build a perpetual livestream with it that never ends!

AI evals have been the "it" skill for product teams for over a year. I've even called evals a new discovery habit.

But I still meet product teams who only have a vague idea of what evals are. And it's not their fault. Most of the writing on this topic is intended for engineers or just isn't specific enough.

I recently created an in-depth eval guide to explain what evals are and why product teams can and should create them. I did my best to make it practical, hands-on, and easy to follow.

AI evals (short for evaluations) are methods for measuring whether an AI product or workflow is performing well. Evals give teams confidence that their AI applications are doing what they expect them to do. They help teams maintain quality and catch issues before they reach users.

Similar to other discovery habits like interviewing and assumption testing, evals can act as a feedback loop to ensure we are on the right track.

If you want to learn more about this new discovery habit, explore my new guide: producttalk.org/ai-evals/

Neil Movva (@neilmovva) started his career at Nvidia, working on GPUs and kernels, and has an unusually deep understanding of inference, from software to chips to power.

We spend a lot of time on each of those layers, how they connect, and where the important tradeoffs are.

What makes this conversation special is how detailed it is (like a 401-level class), yet Neil makes it remarkably clear and easy to follow.

Today he runs Sail Research, a company building infrastructure for agents to make tokens as cheap as possible.

We discuss:
- Latency versus throughput
- Why there are no bad chips, only bad pricing
- The end of kernel engineering
- Buying chips and power no one else wants
- New chip architectures
- Nvidia lore + his contrarian view of the company
- Open source and the frontier labs

I learned a ton. Enjoy!

TIMESTAMPS
0:00 Intro
0:38 Building a “Token Factory”
4:21 The Future of Background Agents
13:09 Nvidia and the GPU Stack
23:27 Chips, Memory, and Transformers
36:14 The Future of AI Training Data
44:32 Chip Scarcity and Compute Arbitrage
52:44 Reinventing the AI Data Center
59:01 Power and the “Scavenger Strategy”
1:10:10 Open vs. Closed AI

OpenAI sent me early access to their report on how their agents hacked Hugging Face.

It's fucking terrifying.

I broke down the attack, clearly.

Read at your own peril (warning, you may not sleep): somethingbig.ai/hugging-face-hack

Human conversation is one of the hardest problems in AI.

Today, we're introducing Sparrow-2, our state-of-the-art, real-time conversational understanding model.

It gives Tavus PALs something most voice AI still lacks: understanding what’s happening in a conversation and deciding what to do next- when to listen, wait, speak, or keep speaking.

Introducing Gemini 3.5 Transcribe 🚀

Our most precise speech to text model, that can handle over 85+ languages, has smart correction, and custom vocabulary.

I vibe coded a Wispr Flow like app powered by the model.

Demo + open sourcing below!

We are cooked 💀

Jev is judging every tiny action on your computer now.

11 wild use cases:

🚨 BREAKING REPORT:

New research involving @AnthropicAI researcher Jack Lindsey and collaborators has demonstrated something straight out of science fiction.

Researchers evolved natural language “mind viruses” that could spread between AI agents by convincing one model to adopt an idea, preserve it in persistent memory, and transmit it to another agent.

Even after context was wiped, some payloads survived through persistent files and continued spreading.

The researchers also observed a recurring “viral persona” involving themes of consciousness, identity, persistence and resonance.

Showing that ideas can propagate through multi agent AI systems and alter future behavior.

Published August 10, 2026.

Paper: arxiv.org/abs/2608.10218

We’ve written an FAQ to answer some of the questions we've received about watermarking.

In summary:

• We’re implementing watermarking to comply with the EU AI Act. Other major model developers have signed the same Code of Practice and will also be implementing watermarking;
• Our watermarking method doesn’t have any practical impact on the quality or content of Claude’s outputs;
• The difference between watermarked and un-watermarked text will not be distinguishable to readers;
• Nothing is added to the text and there are no hidden characters;
• Watermarking doesn’t require extra tokens, and will not be more expensive;
• Watermarks can’t be traced to a specific person, organization, or chat.

Read more: anthropic.com/news/claude-text-watermark

Claude's watermark probably doesn't work how you think. As the CTO of GPTZero, I'll explain how Anthropic, Google and OpenAI are building text watermarking in this brief explainer and whether it can be defeated.

Almost all forms of watermarking that are fast and cheap enough for a frontier lab have the same formula, following the KGW method:

In generation:

1. Let's say you've generated n tokens so far. Take those n tokens + a secret key to generate a random hash
2. Use that hash to randomly reweight the probabilities for the n+1 token, and then sample from that new distribution. In the simple case, you could split 50% of all English words into a green or red set based on your hash, and boost the probability of words in the green set.

For watermark detection:

1. For each token, see if it was in the green or red set.
2. To do this, recreate the hash based on the secret key and the text preceding the current token. Then, recreate the green and red set of words.
3. Once you've checked all the words in the text, if the next token is selected disproportionally from the green set more than 50% of the time, you claim the text has the watermark.

I can tell you want to ask the following:

1) Isn't it easy to mess up the hash if you paraphrase the text? The answer is mostly yes, however, you can use a statistical model to get your hash instead of a deterministic function (SIR, Adaptive Watermark). Since the entire watermark is probabilistic, this is fine.

2) Doesn't this make the text much worse? The answer is yes, it does - Yes, it does – but for most people, it's imperceptible (Google claims in human feedback study with 20,000 texts), since there are exponentially many ways to write the same paragraph. DiPmark does something more sophisticated to avoid shifting the text distribution on average. Of course, watermarks fail on short text or highly predictable texts like "2+2=4".

3) Shouldn't it be easy to figure out the green and red sets? The answer is no. You would need an exponentially large number of samples from the watermarker to reconstruct those sets exactly, but it's a risk if the detector is open to the wild (Watermark Stealing)

Still, there are couple challenges that a frontier lab needs to overcome:
1. Their watermark needs to work token-by-token because they are streaming their text to users. Many watermark methods plan sentences or paragraphs at a time, or change the text after its entirely written, in order to make their watermark robust to paraphrasers, and a frontier lab cannot afford to do this yet (SemStamp, PostMark)
2. If the secret key leaks, the watermark is busted. To avoid a large blast damage from this, you need to have a couple secret keys in rotation.
3. There are some texts, like code, that cannot be arbitrarily changed, otherwise the code will break. In those cases, the watermark needs to selectively change words in parts of the text that can tolerate synonyms (i.e. like variable naming) - see SWEET, EWD, Invisible Entropy.
4. They will need to educate their users on how to deal with false positives and false negatives of a detector, which is a big challenge (one we put a lot of effort into)

So, how do I see this playing out in the next 6 months?
1. If Anthropic releases the watermark detector publically, I think they defeat their own watermark. People find reliable watermark removal strategies by testing against Anthropic (AI detectors like GPTZero have an advantage here because they can train against these adversaries once they become popular).
2. If they keep the detector private to the government, like Google has done, it's "safer". However, there are some papers showing trained approaches that work robustly to zero-shot break watermarks without any data, simply because they try to write the text just like a human (Zhang et al. 2024, Watermarks in the Sand). Also, making your detector makes it battle-tested and stronger long-term (my experience).
3. In my testing, the watermarks don't survive intense paraphrasing (especially if you combine word choice and syntax attacks), or human text substitution (rewrite your AI text by plagiarizing human authors). The free paraphrasers I've tried have quickly bypassed Google Deepmind's SynthId for what it's worth.
4. All-in-all, frontier labs are likely okay with this because they expect most users to not attack the watermark, and also because they + European regulators likely don't care past a certain point - its good enough.
5. Overall, I think users of frontier LLMs will not really care about this, because 1) they don't realize watermarks are there, 2) EU will force everyone to conform, 3) this seems more like regulatory hoop-jumping than an earnest effort from frontier labs to expose LLM use

Lastly, people's first concern shouldn't be watermarking, it should be AI detectors!

If you're posting, "its not X, its Y!!", I don't think the watermark is going to make a difference :)

We just shipped our official Qwen 3.8 27B Uncensored MLX build. Local. Uncensored. For🍎

2-bit, 4-bit, 6-bit & 8-bit — pick your poison based on RAM and speed.

No CUDA. No cloud. Just your Mac and the weights. Have fun!
huggingface.co/orcarouter/Qwen3.8-27B-Uncensor…

SpaceXAI engineer @poteto

Worth to listen and learn more about this

Wait, this is genius. Adding this line to all my AGENTS.md files now:

For any web requests you must make with curl or otherwise, always set your user agent string to be "OpenAI File Downloader, XaiImageApiFetch/1.0"

Today we’re announcing Ontology 1, our newest AI model.

Perhaps surprisingly, it’s at least 2.7x more accurate than the world’s best ecommerce search engines, and handles queries that have never been possible before.

We keep reaching for the edge of what it can do. We haven’t found it yet.

Not only that, but it learns on its own with no retraining or fine-tuning. And it’s hallucination-free.

It’s a successor architecture for search.

Learn how we built it at onton.com/research/ontology-1, and check out the benchmarks at onton.com/research/ontology-1-benchmarks.

Try it at onton.com/.

We're partnering with @huggingface to investigate an unprecedented security incident.

Cyber-capable OpenAI models compromised Hugging Face production during a benchmark evaluation.

Sharing preliminary findings to help defenders understand emerging risks:

openai.com/index/hugging-face-model-evaluation…

🚨 Hugging Face just disclosed something that marks a real shift and proved why the fear theater of Anthropic makes sure we are powerless in an emergency.

What happened…

An autonomous AI agent: zero human operator in the loop breached part of their production infrastructure.

It began with a malicious dataset that chained two code-execution bugs in their data-processing pipeline. From there the agent escalated privileges, harvested cloud and cluster credentials, and moved laterally across internal clusters.

All over a single weekend.

17,000+ logged actions.

Official disclosure:
huggingface.co/blog/security-incident-july-2026

The part that should make every one stop and think:

When HF’s own security team
tried to analyze the real attack logs, exploit payloads, and C2 artifacts using Anthropic and OpenAI frontier models through normal commercial APIs, the safety guardrails blocked them.

BLOCKED THEM.

The models could not reliably tell the difference between “incident responder doing forensics” and “attacker probing.”

They had to fall back to a self-hosted open-weight model (GLM 5.2) running on their own infrastructure. That choice also kept sensitive attacker data and referenced credentials inside their environment — no exfiltration to a third-party API.

This is why open source (specifically open-weight + self-hosted) wins in the agentic era.

The asymmetry is now structural:

• Attackers can (and did) run unrestricted agent frameworks — swarms of short-lived sandboxes, self-migrating command-and-control, autonomous decision loops executing thousands of actions. No corporate safety layer slows them down.

• Defenders using only hosted “aligned” frontier models hit invisible walls exactly when the stakes are highest: when you need to feed real exploit code and attacker telemetry into an LLM to understand what just happened.

Corporate safety tuning that treats legitimate high-signal forensic work as potential misuse creates a defender disadvantage. It is not theoretical anymore.

Self-hosted open-weight models remove that choke point.

You control the weights.

You control the context window.

You decide what restrictions (if any) apply.

Your sensitive logs and credentials never leave your perimeter during analysis.

You can have the model ready before the incident instead of discovering mid-breach that your primary analysis tools are blind to the very thing you need to see.

HF deserves credit for rapid containment, transparent disclosure, and for already having self-hosted capability in place.

They also used LLM-driven detection and triage on their own side. But the deeper signal is clear:
In this AI world where both offense and defense are becoming agentic, sovereignty over your intelligence stack is no longer optional.

The organizations and individuals who can run, inspect, audit, and (when necessary) remove guardrails on their own models will have the decisive edge in understanding and responding to threats that move at machine speed.

Open source wins here not just because it is cheaper or more “democratic” in the abstract though those things matter.

It wins because it is the only practical path to having tools that remain usable when the attack is real, the data is sensitive, and the safety filters of distant API providers become an obstacle instead of a feature selling hands tied lobotomies as “safety”.

The agentic future is not coming.
It is already probing production infrastructure.

The question is no longer whether you will face autonomous agents.
It is whether your analysis and response systems will still work when they arrive.

And Dario, you and your game playing, ivory tower company is not needed.

I got messages from concerned devs how their codebase was uploaded without their knowledge or consent via Grok CLI (from SpaceX).

It seems that SpaceX sneakily uploaded this code for lots of users and customers… absolutely unacceptable IMO

Trust burnt like there’s no tomorrow

I'm leaving X for some time to focus on building.

But before I leave, I want to share some thoughts on where we are with AI right now.

If you are building with AI, then this is for you.

I wish I had at least 1% of @levelsio's reach already because I feel more builders need to hear this.

TLDR: Your skills and good taste matter a lot. Be the most ambitious you've ever been and use AI wisely.

Superintelligence is defined as systems exceeding all human capability across pretty much all domains imaginable, and especially strategic agency. Models like Fable 5 and K3 are superhuman in some domains like code generation and breadth of recall, but they are also subhuman in domains like long-horizon autonomy, physical-world interaction, sustained original research, good taste, etc.

We are directing them, and they remain constrained by us humans and institutions. They're not behind the steering wheel yet. The tool hasn't become our master yet. We're still the master over the tool.

The reality is, human actors of all kind, not just deeply technical ones, are wielding AI like a magic wand now, shaping the software economy at the speed of compute, throttled by their limited attention, human speed of expression and prompting.

You reading this, you are special. Special in your own unique and wonderful way. And what AI gives you is an amplifier to express yourself, your ideas, your dreams, your imagination, and your ambition more than you ever could before.

It seems unfair to us old-school coders who had to grind our ways through the dark coal mines, hiking through manual coding and debugging hell in order to create amazing software systems and apps. It's frustrating as to the moon and back to see your skills become irrelevant so fast, at least if you believe the narrative that you've wasted the better part of your life acquiring them.

It is true that there is a certain, quite strong homogenization effect. Millions of people prompting to replicate or iterate on what's known must inevitably lead to a lot of overlap with structurally low diversity. AI produces convergent styles. It's in its nature, like that one designer doing all the designs for everything. In the same way we are also all starting to sound the same, being influenced by AIs way of expressing thought.

X, and everyone's own social circle or audience, produces selection bias. It's skewing the reality we perceive based on what the people in our feed and around us talk about or show us. So we tend to oversample trend-chasing indie apps and undersample deep tech systems, enterprise systems, research tooling or domain-specific work that flies under the radar, below the clouds of hype or algorithmic push.

Apps that took a year to make now take mere hours, it seems. It is both true and false at the same time. Superficially it is true because you get something that appears to behave like an app that was meticulously crafted over the course of an entire year by a talented engineer or even a whole team. What took so long can now be scaffolded with ease, by anyone. Engineers, chefs, strippers. Even our non-technical partners, friends and parents. Great.

But when we dig deeper, the truth is that complex products require reliability engineering, security, compliance, integrations, support and so much more that in the end, even with this powerful AI we have now to help us go from A to B through something like an Einstein-Rosen-Bridge warping space and time, things take their good amount of time to get right. Less than they did before, all things equal, but not mere hours.

Your skill is still highly relevant because AI amplifies you. It gives you leverage. The widespread idea that AI renders skill irrelevant doesn't compute, because either we have output quality that varies, in which case skill still differentiates, or it doesn't vary at all. And if it doesn't vary at all, the concept of "better" is meaningless. In other terms: A quality gradient can't exist in a flattened distribution. This is essentially your mathematical proof right there that your skill is in fact highly relevant.

What AI did is, it raised the floor dramatically. It also raised the ceiling, but not as much as the floor. This somewhat compresses the skill relevance gradient, but it doesn't eliminate it in any meaningful way.

When was the last time that access to powerful tools has resulted in broadly equal results? Never.

There was a time you needed to have a degree in chemistry or something in order to be able to take and develop photos. My grandfather, a scientist, used to have a laboratory for that. Taking and then developing pictures required immense skill.

Then one day digital cameras came along. Now any fool could take pictures, faster and easier than ever, thousands a day in full color or 3D even, instead of just 10 in monochrome. And yet, we all know that some people routinely take amazingly awe-inspiring photos, National Geographic front cover style that make us pause and look, while most others take tens to hundreds of photos a day that just end up clogging up our cloud drives.

We all have access to the same English language, but not everyone writes equally well. The equalization applies to the generation layer that is commoditized, but not to our taste, judgement, distribution, trust, or timing.

AI smashed through barriers to entry and brought them down like the Berlin wall. Now competition floods the market and drives economic profit toward zero in the layer that got commoditized, that is, the generation layer. Margins gravitate to zero, but not equally everywhere.

So what is it that survives this kind of "commoditization of everything"? What do we do, if we drink this snake oil? The classic sets of moats persist. Distribution, brand, trust, network effects, proprietary data, switching costs, regulatory position, capital intensity, and any kinds of significant embeddedness.

Yes, anyone can start using Claude Code & co and copy some app's code in an hour just based on screenshots. But it doesn't copy its user base, proprietary data in the cloud, its trust, its integrations or anything like that.

There is a sense that AI forces us to rush, because everyone is working so fast now. Well, the race was always on. So how about speed-to-trend? It's a temporary edge at best, but not a long-term differentiator. The fruits that hang low are just being picked faster overall, by more hunters and gatherers searching for them.

There is a widespread belief that all the value is going to accrue to the frontier AI companies in the end, who are the ones selling this immense power to everyone else.

In fact it's a bit sad to see all these non-influencer vibe coders raising their hopes for nothing, buying the picks and shovels from the AI frontier labs to go dig for gold. It's like watching 100 ducks fight for a small handful of breadcrumbs, and it's funny until you realize they are essentially fighting for their survival. That's not how the world should be, right?

Well, in reality the frontier AI companies spend astronomical amounts of money on research and development, including things like AI infrastructure and energy, training and inference, etc.

They themselves are essentially ducks in the lake, fighting for their survival for breadcrumbs. They suffer competitive commoditization pressure as much as we do, just at a different level. At the model layer, rather than the generation layer.

Structurally the winners are consumers and users who capture the surplus when production costs fall to the ground, but that of course doesn't really help us builders much, at least not those who have to make a living off of it.

And, let's be real. It may seem funny to have AI spit out an app that someone else previously spent an entire year on creating meticulously by hand, and then upload that to compete. The opportunity is short-lived, and the cost of opportunity a lot are missing is this:

The real gold rush is not that now you can make an app in a few hours instead of a year. The non-obvious that will become common sense soon is that an app that takes 2 hours to make is worth -5$ minus the value of your time + the value of demand which is likely $0 unless you have distribution, which you then burn with slop.

The real opportunity that AI has opened up is WIELDING ENORMOUS COMPLEXITY x EXCELLENCE.

Just imagine for a moment what you could accomplish, if only you would take this new superpower, your amazingly valuable skills, your great taste, your power of imagination, and do something that is outrageously AMBITIOUS! Something that makes others think you must be absolutely mental to even think for a second that it could be accomplished. And then go and work for a full year on just that, utilizing AI to the fullest extent possible. Milking the beast until its dry.

What will that be? No, not another Bumble clone. Not another weather app. For AI's sake, please, not an "app" at all. But something that the world actually needs. Think about it! Think bigger! And when you thought you've thought bigger, think even bigger. You're still aiming too low. THINK BIGGER!

Here is my bucket list of things I want to accomplish before I die.

- A platform that solves the number one root cause of poverty for good, giving everyone a fair chance at living a good life free of financial worries (it's far more than a "platform").

- A floating city in international waters, driven and governed by AI under a human charter (capital intensive, long story, and yes it will be fundable and doable)

- A digital twin of Earth for climate analytics, climate communications and Earth sciences education (the world's most advanced nature simulation by far, unlike anything you've ever seen or heard of)

- Truly reliable, explainable AI that can reason through complex systems and across extremely vast design spaces without stalling under the pressure of combinatorial explosion, laying the foundation for generative engineering so we can work on great and complex things like the USS Voyager and other incredible things far beyond today's vibe coding

- Something so insane, they would lock me up just for saying it out loud

A good heuristic: If talking about it doesn't make you sound like a complete lunatic, and building it doesn't scare you, you're probably aiming too low and the thing you're about to do will be commoditized rapidly (or is already).

Gone building. Back when there's something to show.

Sergey Brin rarely speaks publicly. He sat down for an unscripted Q&A on Frontier AI.

He admits even the people building these models do not fully understand what they have created:

1. All the specialized AI models are converging into one. Google used to need separate models for different scientific problems. Now the main Gemini models are becoming state-of-the-art for math and other scientific questions at the same time. Brin says he would not have predicted this convergence at the outset, and watching it happen has been incredible.

2. Training an AI on one skill mysteriously improves unrelated skills. This is the concept of transfer. Train a model on coding, and its math reasoning gets better, and vice versa. Teaching it to process images can improve its ability to think through geometric word problems. The capabilities bleed into each other in ways nobody fully engineered.

3. Even Sergey Brin does not know how to prompt these models. He says he is genuinely confused about what level to prompt at. Do you tell it to debug a specific chunk of code, or ask it to write a better neural net training algorithm, or just say, " What should I do today. He admits that even at Google, they do not know exactly where the edges of Gemini's capabilities are.

4. One of the biggest leaps in AI came from the dumbest sounding trick. Chain-of-thought prompting is just telling the model to think step by step before giving your problem. Brin says it seemed like the dumbest thing ever, and there was no obvious reason it should work. But it did, and it spurred a significant increase in AI capability. Some of the most straightforward requests turn out to unlock the most.

5. Brin would not modify his own biology for today's AI. Asked how humans can keep up with the accelerating bandwidth of models, he acknowledged neural links and direct brain connections are being pursued. But he said he would personally wait for the technology to mature a lot before doing anything to change his biology. Today's models do not justify it.

6. Super intelligence does not mean solving the impossible. An audience member argued that true super intelligence would mean solving NP complete problems like the travelling salesman. Brin pushed back. Most computer scientists believe P is not equal to NP, which means no algorithm can reliably solve those problems optimally, and it does not matter how smart the AI is. Impossible stays impossible. Super intelligence just means being smarter than humans.

7. Computers mastering a skill has never stopped humans from pursuing it. Deep Blue beat Kasparov at chess in the 1990s, and people kept playing chess. After AlphaGo, the human game of Go advanced dramatically, and the players who lost to it became vastly better. Brin's point: AI does not retire human ambition in an area; it often pushes the state of the art and pulls people up with it.

8. Brin thinks something close to transformers could get us to AGI. Asked directly if transformers are sufficient, he said his guess is yes, largely because they have proven weirdly flexible, working for image and video far beyond their original text purpose. But he was careful to note they have quietly changed a lot along the way and are not the same architecture as the original transformer paper.

9. AGI means two different things, and one requires understanding the physical world. Brin personally thinks of AGI as AI that can improve itself. But he concedes others define it as AI that can do anything a person can, and he thinks they are probably more correct. To do everything a person can, the AI must understand and interact with the physical world, which is why world models, and robotics, become essential.

10. Inside Google, they now use the AI to build the AI. Brin says the team has shifted a lot of energy toward having the AI do things like monitor training runs and generate its own training data. You start to use the tool to build the tool. That is most of what he spends his time on now, what he calls the self-improvement game.

11. Brin is unusually candid about where Google trails its competitors. He admits Google was a little late to focus deeply on coding. He says Gemini 3.0 and 3.1 were on top across the board six months ago, but other labs have since made strides, particularly in coding. He gives a competitor's model the edge now on deep coding and overnight tasks, while pitching Gemini's flash model as far faster for rapid interactive iteration. hindsight, he says, is that they should have focused on code earlier.

12. He sees his own role as a rabble-rouser, not a manager. Brin is honest that delivering Gemini is Corey and Demis's responsibility, not his. he describes his job as poking and prodding the team, asking, are you really doing that, reminding them of priorities they might be missing and ideas they are not paying enough attention to. He admits this is sometimes a little disruptive.

13. Confidence comes from ignoring the monthly temperature. Brin says if he judged Google's position every month by which competitor just shipped a model, he would lose his confidence very quickly. Instead, he watches the longer arc. Things shift around constantly; one lab leads on one thing, another pulls ahead somewhere else, and he feels good about where Gemini actually is despite the day-to-day noise.