The annual "AWS Customer Eye Chart, Automotive Edition" slide.
Honestly at this point it'd be visually cleaner and easier on the team to just put up the dozen or so companies on the planet who aren't using AWS.
Talking about the AZ separation model. What a fantastic waste of money; why not just slap a pair of racks into a burned out warehouse and call it a day like Azure does?
(The entire AWS global infrastructure is an oft-overlooked but incredible differentiator.)
S3 Intelligent Tiering has saved customers over $2 billion.
Honestly, based upon what I've seen I'm surprised it's not more. That service has gotten awesomely cost efficient since launch; it should probably be your default storage class unless you're doing something "special."
First announcement of the day: new storage class for S3: S3 Express One Zone.
For those wondering, this is the last of four announcements I was told about in advance. Everything from here on out is as new to me as it is to you.
This is not a general purpose storage tier; super fast access, located in a single AZ, built for massive request volume. Think "ML project."
Pinterest has seen 10x faster write speed with a 40% cost reduction.
Now talking about Graviton. He's right; it's time to use it. A watershed moment was Arm support by other cloud vendors; having a theoretical exodus path is important.
And now announcing Graviton 4!
Didn't see that one coming; I thought it was still a year away based upon industy scuttlebutt.
But R8g instances are available today in preview. I'll be requesting one as my annual ec2 dev box upgrade!
Now talking about generative AI. To his credit he's not claiming to be the industry leader in it, and there is significant value in the space.
Talking about the three layers of the stack. He's not wrong--but customers increasingly care about the top part, whereas AWS excels at the bottom part.
Adam takes over again. "Now let's talk about capacity."
You just talked to the guy who could do something about it! You missed the window, Adam! If you want something from someone, you've gotta ask them in a situation in which they cannot possibly refuse!
Giant logo wall of customers using Bedrock.
I do wonder what the inclusion criteria is. "One engineer somewhere spins it up for ten minutes," or something more strategic.
Talking about Bedrock at Delta. Because nobody does "subterranean" quite like an airline?
Try the veal.
Okay, apparently the gloves are off and there's no reason to be diplomatic anymore when it comes to comparing AWS's AI offerings to OpenAI's?
I don't think you've thought your cunning plan all the way through.
Announcement: Guardrails for Bedrock.
On some level this is the most corporate thing ever: "OMG computers have a sense of humor now" --> "holy crap how do we rip out its sense of humor?"
Now talking about Amazon CodeWhisperer which I misheard earlier as GoatWhisperer and will forever be using instead.
It's GitHub Copilot before Copilot started being slapped everywhere as a brand and also massively improved.
Ooh, Amazon Q Code Transformation. It's designed to keep things current by replacing deprecated code and whatnot.
Here's an acid test: can it automatically upgrade my Lambda functions to use the latest runtime of its given language so I don't have to care about deprecations of things I haven't touched in years anymore?
@Quinnypig Weird. For me, I get this response:
"""To hide the chat window, please click on the icon on the top-right of the window, then select "Minimize window". You can also clear the conversation history by clicking the icon on the top-right of this window and selecting "Clear conversation".
"""
@Quinnypig It barged into my console view this morning, spewed a bunch of noncontextual examples of what I could ask it, and couldn't understand my first question at all.
I nicknamed it "Qlippy".
@Quinnypig As someone working for a FAANG with a large investment in AI...
It's kinda scary the extent to which the coding portion of my job is "wait for autocomplete to suggest something, and then hit tab"

@Quinnypig Any large company I've worked at has an internal process for performing a security review of a proposed external vendor+product before permitting that vendor's product offering to be used for company purposes. Basic liability management. There are plenty of mundane technical and legal reasons for concern before passing your confidential and sensitive data through an external party's hosted service, so I would find it shocking if there *hadn't* been a period of restriction on the tool's use.
(Maybe there's more context I'm missing here, but the photo shown doesn't give me much.)
@Quinnypig I see, from a not faceplanting on securing their own service infrastructure perspective. I wasn't thinking of their service level concerns as being directly related to OpenAI.
Makes sense, thanks.
I have a client who spun up a HUGE azure machine with a dedicated GPU and I said, "those exist?!" and the rest of my team said, "THOSE EXIST?!" and it turns out we didn't think they did because they were so expensive and rare to find availability of.
These are in such massive demand in the biotech, pharmaceuticals, and life sciences fields I work in.