Matt Asay
@mjasay
Followers
23K
Following
19K
Media
2K
Statuses
46K
VP, Developer Platform @Oracle. Formerly @MongoDB, @AWScloud, @Adobe. Arsenal apologist. Weekly columnist for InfoWorld. Daily skier.
Salt Lake City, Utah
Joined July 2007
Richmond “gets” agent memory like few others. I channeled him a few weeks back for InfoWorld on the topic:
infoworld.com
The humble database offers the key to giving AI context and adaptation, accessing data beyond its training cutoff.
𝐃𝐚𝐲 𝟐𝟕/𝟏𝟎𝟎 𝐨𝐟 𝐀𝐠𝐞𝐧𝐭 𝐌𝐞𝐦𝐨𝐫𝐲 🧠. That Was Absolutely Insane.4 Hours Live Training session on Agent Memory.Thanks for the support @OReillyMedia . I just finished probably one of the longest teaching sessions I've ever done. I haven't slept in 24 hours, and I've
0
0
1
I discussed this with Richard a ways back: what he and the team do with docs is magical. Really smart, raise-the-bar, deeply impactful work in an area that others treat as unimportant 'if I have time' sort of work.
My team covers developer relations and documentation for @googlecloud. How are we disrupting *ourselves* with AI? All sorts of ways!. Helen and Matt are two of our leaders changing how work is done. Here's a peek at a couple key efforts:
1
1
5
I have loved being able to run Deep Research projects (I use it all the time), but now I have no sense for when I'm "getting my money's worth," as it were, and when I'm just benefiting from the free version of ChatGPT.
0
0
0
I *think* I still have Deep Research access,but there's no longer any way to see it in the UI. I can see from here ( that I have access (but even "deeper" research if I were to go Pro, if only I had a way of knowing what that means .
help.openai.com
Information about our paid subscription plan, Pro.
1
0
0
OpenAI/others have been under pressure to "productize" GPTs (something @benedictevans has written cogently abt , but GPT 5 is a step backward IMO. I used to know ChatGPT Pro got me x number of Deep Research queries, but now I've no idea what I'm paying for.
ben-evans.com
We’ve had ChatGPT for 18 months, but what’s it for? What are the use-cases? Why isn’t it useful for everyone, right now? Do Large Language Models become universal tools that can do ‘any’ task, or do...
1
0
1
The backbone of open source isn’t shiny launches or Twitter fights—it’s boring, essential maintenance. And the folks doing it aren’t who you think. Credit the quiet contributors driven by self-interest, not halo polish. 👇 . My column for @infoworld
infoworld.com
We all depend on open source basics like the Linux kernel, Kubernetes, or security and maintenance software. Let’s take a moment to acknowledge the people—and companies—keeping them in shape.
2
0
3
A reminder that, yes, these new models are awesome (hey, they're open source, so obviously great, right? 🧐), but they still hallucinate, still require lots of care and feeding to make them useful, etc.
It's a real shame there are no LLM benchmarks on hallucinating creativity, since GPT OSS 120b is wild in what it invents out of thin air, and would score really high on such a benchmark.
1
0
2
84% of developers use AI, but only 33% trust its output. Why? Bc most AI‑generated code is “almost right”—and that gap costs time, trust, and production‑readiness. Hence, developers' role changes to be AI supervisors, validators, and orchestrators
infoworld.com
If you want trustworthy AI results, you need trustworthy people shaping the prompts, verifying the data, and overseeing the whole AI process.
0
0
4
🤔 "Oracle Databases are widely adopted, and the new MCP server will accelerate enterprise AI integration by allowing enterprises to automate Oracle Database interactions based on natural language queries that activate agents and their supporting LLMs."
0
0
0
AI doesn’t need more parameters—it needs more *memory.* Today’s AI apps are essentially amnesiacs, but. plug in a database and you get context, recall, and adaptation. I walk through the agentic memory stack in this week's @InfoWorld column
infoworld.com
The humble database offers the key to giving AI context and adaptation, accessing data beyond its training cutoff.
1
0
3
"Every vibe-coder is now generating as much technical debt as 10 regular developers in half the time." 😳😵💫😭.
A friend asked me to look at their code, and *Holy Mother of God!*. He works for a medium-sized company. They are building some sort of internal CRM. They've been vibe-coding some of the new features. They are professional developers. Most, I assume, know what they are doing,.
3
1
11
RT @GergelyOrosz: I am getting tired from influencers with little to no tech industry experience to write stupid stuff like this for likes.….
0
330
0
One nice thing about joining Oracle is it gives me a regular opportunity to go back home to Stanford. Is there a better campus on this planet? I don't think so.
1
0
2
Great tech. Great CEO. Great @antirez. Lots of great!.
The AI and Agentic wave is here! We're growing at Redis and we're looking for the best people to join. For example: VP of Demand Gen. Come cook with us.
0
0
1
I'd not considered this, but Zack is right. AWS has *tremendous* internal assets that will feed an LLM beast (for analyzing strategy, authoring 6-pagers, etc.).
AWS has struggled thus far in terms of customer-facing AI, but I suspect there are few better positioned to use AI internally. The corpus of structured internal writing is incredible training data – PRFAQs, design documents, 6-pagers – & docs are actually used to drive products.
1
0
1
Speaking of transfers to other teams. I'm very excited by this addition to Arsenal's. arsenal. :-).
🚨❤️🤍 BREAKING: Viktor Gyökeres to Arsenal, here we go! Verbal agreement in place between all parties involved. Sporting accept last bid from Arsenal for €63.5m plus €10m, agent will reduce his commission. Gyökeres will sign five year deal at #AFC. He ONLY wanted Arsenal. 🇸🇪
0
0
0
Tl;dr? I just left MongoDB to join @Oracle. Not yet ready to share all the details but it involves my two favorite things: 1/ helping organizations make tough cultural changes while 2/ ensuring developers can build the best possible applications .
linkedin.com
Over the arc of my career, I’ve been called many names (some favorites: “open source zealot” by Tim O'Reilly, cloud apologist while working at AWS...), but I’ve come to be very comfortable as a...
8
0
30
LLMs don’t follow scripts—they improvise. That’s why your old-school authorization model won’t cut it. We need security that travels with the AI. I break it down in my latest @infoworld column:.🔐
infoworld.com
Give a large language model broad access to data and it becomes the perfect insider threat, operating at machine speed and without human judgment.
1
2
3
And I was there with my wife, watching that magical match with the Arsenal fans. So. Much. Fun.
The guy at the back with the beer glass, dark blue cap and glasses on once scored four goals at Anfield against Liverpool. 🍺
0
0
1