
Engineering @ Chainguard ⛓️ • Book Author • ex-Elastic ($ESTC) • tweets are my own
US - Boston
Joined March 2009
- Tweets4.8K
- Following1.3K
- Followers804
- Likes4.5K
Maxime Gréau retweeted
Bugs are getting exploited faster than ever, and most teams still aren’t using the strongest, most capable models to find them first.
Chainguard's harness builds your app from source in an isolated sandbox, then points the latest frontier models from Anthropic, OpenAI, and Google at it to hunt for exploits in your first- and third-party code. When it finds one, it verifies the exploit and hands you a verified fix to merge.
Chainguard Chief Product Officer @prdonahue shares more.
Join the waitlist: chainguard.dev/products/harn…
Maxime Gréau retweeted
Regardless of what it’s called, we’re clearly moving to yet another era of AI agents in our work.
We’ve quickly jumped from agents that spit out a quick answer for you, to today agents that you orchestrate inside of a set of tasks. But as agents can both fan out and work in parallel, handle much larger context, be able to do long running work, and have far greater levels of capabilities, we’re entering yet another level of what we can have AI do for us.
We’ll soon be able to give agents a set of goals, and they can go out and design the best solution for that goal, while also continuously improving it based on the ongoing execution of whatever it built. This will obviously start with engineering because that’s where we’re seeing the best practices emerge first in almost every new era of agents.
But the impact can be quite massive across the rest of knowledge work. You’ll tell an agent to improve your client onboarding workflow to achieve a certain targeted goal; or have agents go reduce risk in your business; or improve your security posture on an ongoing basis.
These autonomous systems will be able to take on a set of goals, design the best outcome, test and iterate, deploy, and then optimize whatever it built on an ongoing basis. Importantly, this leads to outcomes and results that humans literally couldn’t have accomplished before. This will lead to fairly new levels of productivity with AI.
I built an agents board 2+ months ago, agents-board.mgreau.dev/
it was obvious that was coming after the OpenAI / HuggingFace story
Introducing 𝚟𝚎𝚛𝚌𝚎𝚕 𝚋𝚘𝚊𝚛𝚍, a secure message board for people & agents.
• Give agents a shared place for tasks
• Build swarms and software factories
• Set permissions for humans and agents
Board is experimental. Tell us what you think. vercel.com/blog/introducing-…
1
3
158
Maxime Gréau retweeted
The folks at @calif_io beat us again by a few weeks, we found and reported java4shell independently Friday night while scanning an open source Java project.
We'll beat them soon.
chainguard.dev/unchained/jav…
Maxime Gréau retweeted
What @reedhastings understood — and what I didn’t yet — was that one of your most important jobs as a founder is to work yourself out of every job in the company.
One by one.
Maxime Gréau retweeted
Today, I published my analysis of @AnthropicAI’s S-1, tracing the investment case, the assumptions behind my valuation, and the risks that could change my view.
$ANT officially launched its roadshow today and appears to be targeting a valuation between $1.2T to $1.4T.
My own numbers sit above that range.
If everything an investor has to believe about the company holds, the equity is worth about $2.24 trillion before the offering. Once I price the ten events (Single Points of Failure) that could break the case even if those beliefs hold, it is worth about $1.51 trillion.
The stakes surrounding this listing extend far beyond Anthropic.
A listing this size gives public investors something they have not had before: the filed accounts of a U.S. frontier lab. And because Anthropic sits at the heart of an ecosystem of financial partnerships, the future of a vast array of companies is now tied to its success or failure.
And yet, analyzing this company presents a unique challenge. Anthropic is something we haven't seen before. It's a software company with the balance sheet of a utility.
I think Anthropic has a credible path to becoming one of the defining companies of this cycle. But 2027 is going to be a major test, as compute obligations put intense pressure on revenue growth to pay the bills.
In the latest episode of the AGNT Podcast with @GemmaAllenSays, we discussed my views on how to think about Anthropic's valuation, its revenues, compute challenges, financing, and how to measure the right benchmarks.
Read the full analysis from Decoding Discontinuity and download the 30-page teardown here: decodingdiscontinuity.com/p/…
cc: @furrier @bjbaumann2014 @NYSE @NYSEWired @SiliconANGLE
Maxime Gréau retweeted
Palo Alto Networks CEO @nikesharora predicts AI is going to fundamentally change how companies are structured:
“A lot of organizations are structured in silos. You have documentation people, testing people, coding people, product people, UI people. You have 8 different sets of people who combine together to build a product, which makes no sense.”
“The only reason for this [way of working] is because nobody could be amazing at everything.”
“But if you believe that you can harness all this intelligence in a way that one person could master all these 8 crafts because they have very able assistants which will do 90% of the work for them, it fundamentally changes the way we get shit done, the way we build things, the way we do anything in our lives, the way companies are created.”
“The shape of a company will be fundamentally different 5 years from now.”
.@NikeshArora grew Palo Alto Networks from $18 billion to $250+ billion in market value. Before that, he spent 10 years helping grow Google’s revenue from $2 billion to $62 billion.
When he joined Palo Alto, he thought cybersecurity was two words. 8 years and 35 acquisitions later, the company has more than 70 products.
His approach: buy the winning team and put its founder in charge. If a startup beat his team with fewer resources, why make its founder report to the person who lost?
Nikesh explains why he makes founders agree to a product roadmap before signing a deal, why he wanted Google to buy Netflix, and how Egon Durban kept him bidding for a cricket team at 6 in the morning.
0:00 The Oculus deal he almost did with Egon Durban
8:45 "I thought cybersecurity was two words"
11:44 Why he buys instead of builds
14:33 Why the acquired founder becomes the boss
16:21 Why he only buys number one or number two
19:30 The 30-day test before every acquisition
33:38 Egon Durban and the 29-round cricket auction
42:32 Why nothing at work stresses him out
44:28 Selling Google ads without knowing the price
49:42 Coming to America with $200
51:58 Raising kids who didn't grow up poor
54:46 Why AI hacking has been good for Palo Alto
Includes paid partnerships.
Maxime Gréau retweeted
open.spotify.com/episode/1eD…
@davidsenra thank you for coming over to do this. Was fun.
I really don’t understand how Google expects me to use Gemini across its products. The gap between what others are doing with Anthropic or Ramp products and what Google offers is astonishing.
I’ve tried Gemini several times across multiple Google apps, and it’s always the same story: it feels like using a chatbot from 2022!
50
🎯 Building custom interfaces with a prompt via MCP, as well as through the UI, is going to become the default way to share the right information with your audience (engineering teams, customers, C-level executives, etc)
73
💯 I’ve not executed any terraform command manually in 2026 to deploy and test changes in my GCP and AWS dev env, my agents figure all it out thanks to IaC foundation put into place years before AI
IaC isn't dying, it's becoming more essential. Non-deterministic layers add new volatility to the stack, but they don't break the core principles. For example, it's becoming more obvious that we want some layers to be truly reproducible.
2
75
Maxime Gréau retweeted
Replying to @GergelyOrosz
@GergelyOrosz is right. From my pov there are two reasons to write your own stuff, even if it's a struggle. First, it's fitness training for your brain. If you struggle to write, your brain is flabby. Having a fit brain will be a key differentiator in the new era. And second, we would all rather read your "bad" writing than polished AI writing. Trust me on this one. We can spot even one sentence of AI writing and it stinks. Your bad writing is fine. Just do more of it, every day, to get in shape and stay in shape.
Made the previously paid article "Becoming a better writer" free.
I see too many devs writing low quality slop with AI. Hopefully this helps those of you who want to share more of your own experiences, using your own words - we need more of these!
newsletter.pragmaticengineer…
💯 A major shift toward proactive, always-on, goal-oriented agents is starting now and will gain momentum over the next few months
Maxime Gréau retweeted
so now that you have entire teams of agents running your company...
♫ what do we do while your agent runs? ♫
Maxime Gréau retweeted
ai has made many new things possible
how should an org deal with it? what should you do?
i got sick of the long slop docs, shared videos and unusable apps
sent this to the teams inside @Snowflake
warning! actual sentences follow.
--------
Always, always look at what you are suggesting from the perspective of the recipient. Empathy is the top AI skill.
Information:
-- Aim to inform; concisely; Nobody likes reading reams of material
-- Always add value: don’t share a video without doing the work to generate a verified summary and / or your perspective.
-- Don’t flood people: a 90m video to watch is 89m too long
Apps:
-- Always start from your end user perspective: what do they want? Think like them.
-- Meet users where they are: show up in email or slack if you can. Cowork / coco are the next options. Unless for a specific high volume task, nobody wants to use another app. They are impossible to find and generally really hard to use.
Don’t confuse “AI native” for “flood of new apps” or “flood of documents”. AI native is about reimagining work to be simpler & easier allowing us humans to focus more on judgement and rumination.

























