Sign in to view Ryan’s full profile
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Sign in to view Ryan’s full profile
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
San Francisco, California, United States
Sign in to view Ryan’s full profile
Ryan can introduce you to 10+ people at Microsoft AI
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
2K followers
500+ connections
Sign in to view Ryan’s full profile
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
View mutual connections with Ryan
Ryan can introduce you to 10+ people at Microsoft AI
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
View mutual connections with Ryan
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Sign in to view Ryan’s full profile
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
About
Welcome back
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
New to LinkedIn? Join now
Services
Activity
2K followers
-
Ryan King posted thisThe transition to agentic coding is apparently a slow process of the industry learning that Amdahl’s law applies to organizations, not just running software.
-
Ryan King reposted thisToday makes not one, not two, but three top-tier models shipped from Microsoft AI in just the last few months. Announcing MAI-Transcribe-1 with state-of-the-art transcription, joining MAI-Voice-1 and MAI-Voice-2 in our model suite. We've already been shipping them into production - now you can too, as we expand access on Microsoft Foundry. Meet the trifecta 🔽 MAI-Transcribe-1: the most accurate transcription model in the world across 25 languages, according to industry-standard FLEURS word error rate benchmark. MAI-Image-2: debuted as a top 3 model family on arena.ai and hailed as "a genuine game-changer" by Rob Reilly, Global Chief Creative Officer at WPP, one of our early adopters. MAI-Voice-1: sets a new bar for natural and expressive speech and can generate 60 seconds of audio in just 1 second. We're already shipping them into production across Microsoft - now you can build with them, too. All three available today for developers on Microsoft Foundry, and you can try them for yourself at MAI Playground (US only): https://lnkd.in/gHyE2ZfY Better, faster, and with killer price-to-performance. More on today's blog: https://lnkd.in/g2BFj3Z5
-
Ryan King shared thisMember of Technical Staff - Developer Experience | Microsoft AIMember of Technical Staff - Developer Experience | Microsoft AI
-
Ryan King posted thisHey folks, I took a new job. A few weeks ago I started a new gig on the Microsoft SuperIntelligence team. I am initially going to be focusing on developer experience for our team of engineers and researchers. If you are interested in working on AI or with tools for python, c++, cuda, gpus, azure, etc, let me know. I am sure we have a spot for you on the team.
-
Ryan King posted thisBeen somewhat quiet on here lately and here is why.. The last few years working with Cruise and Cruise+GM have been... interesting, but came to and end for me over the summer. After taking a few months to recharge and have some family time, I am on the hunt for my next adventure. If we are not familiar, I have 20 years of experience, working mostly on technical + human systems problems related to scaling services and teams. I prefer to work as an individual contributor, but have years of experience as a manager as well. I would be happy to talk to anyone about roles for which I might be a good fit.
-
Ryan King posted thisI have been a bit out of the coding game for awhile (more on that soon...), so I have been a bit skeptical of of the hype around AI-driven software engineering, but also haven't gotten a chance to really use it freely. I was very worried that it would take all the fun out of coding, but I am so far finding the opposite. I have an army of cheap interns who don't complain about grunt work. No more grunt work for me (for any class of grunt work I can describe).
-
Ryan King reposted thisRyan King reposted thisTwitter almost ran out of IDs. Six weeks until it would stop accepting new tweets. Full stop. Ryan King told me what that felt like from the inside. A public countdown. Third-party apps that relied on the ID format. But they fixed it, buying themselves another bit, from 31 to 32 bits. Then it happened again. Another 6 weeks to migrate everything, this time from 32 to 64 bits. That was harder. Then again: every tweet lived on a single machine. They sketched a plan in a week, told everyone immediately, and then raced the timer. The save wasn’t pretty: a three-week hack that shoved new tweets into a fresh partition and let reads walk backward in time. They shipped it dark, flipped a single pointer, and it worked. The “temporary” fix ran for almost two years. That bought the space to go truly distributed: build Snowflake for IDs, discover the JavaScript 2^53 landmine the hard way, keep moving. With the fire out, they built the "right" thing, on Cassandra. 8 months in, the cancelled the project. If these problems feel different, it's because they do. It's not just the scale. Twitter's solutions changes the industry forever. The whole interview is amazing. I'll drop a link in the comments.
-
Ryan King shared thisI have really enjoyed my time at Cruise, and currently still have a job, but with this news I have to say I am open to talking to folks looking for Principal Engineering type folks. https://lnkd.in/gyFkv3ceGM to refocus autonomous driving development on personal vehiclesGM to refocus autonomous driving development on personal vehicles
-
Ryan King reposted thisRemote Assistance is the primary mitigation when AV can't autonomously navigate challenging and long-tail scenarios. 🌈 Sarah King is an amazing leader and we are looking for engineering managers and a test engineer to join Remote Assistance team.Ryan King reposted thisI’m #hiring. Know anyone who might be interested? * Engineering Manager, Fullstack, Remote Assistance * Tech Lead Manager, Remote Assistance AV Engineering * Senior AV Test Engineer, Remote Assistance
-
Ryan King liked thisRyan King liked thisI'm so very confused by all the people who talk about how many tokens they're spending, how long their sessions work, how many different tools they're using simultaneously... and don't actually describe what the business result is that they're actually achieving. Impact >>> effort.
-
Ryan King liked thisRyan King liked thisOm Malik passed away, and I find myself thinking of the small things. I was a New Yorker living in Silicon Valley in those years, which is its own particular kind of displacement. Om was a New Yorker too. He had traveled a long road to get there via London before the pull of the technology story brought him west. Somewhere along that journey, through cricket and then through those New York years, he had landed on the Yankees as his team and also explained cricket to me. In a world of relentless West Coast optimism, two New Yorkers finding each other meant something. I took Om to baseball games. And somewhere between innings, in the way that only that setting allows, we would have the real conversations. Om talked about what it meant to grow up inside the Silicon Valley tech scene as a person of color, how to punch through, how to find your footing, how to live inside the bubble of that moment without letting it go to your head. He talked about the people the industry was leaving behind, the lives being disrupted by all this tech we were so proud of, the empathy that was missing from the room. He was a north star for me in that way, one of the few people of color then whose developing voice and presence carried weight for me. I came to Silicon Valley through journalism myself. I was working at Knight Ridder Digital, at the moment when the newspaper giant was finding their footing in digital journalism, and I understood from the inside what it meant that Om had built an independent voice unattached to legacy media and had made it matter anyway. He was conscious of what that meant as a person of color in that space. I didn't know it then, but he was already working out the argument he would make in a 2016 piece for the New Yorker. That Silicon Valley's greatest failing wasn't its products or its promises, but its distinct lack of empathy for the people whose lives it was upending. I heard that argument first between innings, as a private conversation, years before the rest of the world did. There was one evening when we had all gone to dinner and came back to find my car had been broken into, windows smashed. Om helped me take stock of the damage, and once we realized nothing had been taken, he got very matter-of-fact about the whole thing. He looked at me and said, "I think they just wanted to hang out." That was Om. He had a way of making the world feel a little less alarming than it actually was. When I moved back to New York, we had one of our last conversations. It struck me then, and strikes me now, that we were all just trying to figure out who we were going to become. He became someone the whole industry knew. I'm glad I knew him when it was still just baseball and broken car windows and the kind of advice that stays with you for twenty years. There will be no shortage of tributes to Om Malik the journalist and the institution-builder. I just wanted to add this one small corner of him.
-
Ryan King liked thisRyan King liked thisThe longer I work with AI, the more convinced I am of one thing: it rewards taste. The better your taste, the better the output. So what is "taste" in software development? It's knowing what to ask for. Knowing what onions, hexagons, and mermaids have to do with architecture. Knowing what ADR, DRY, YAGNI, and SOLID actually mean and when they matter. That's why the AI divide is so stark. Fastly surveyed 791 developers and found that senior engineers ship ~2.5× more AI-generated code than juniors. A third of seniors say half their shipped code is AI-generated — vs just 13% of juniors. The reason isn't access. Everyone has the same tools. It's that seniors can tell when AI's output looks right but isn't. AI amplifies the judgment you already have. So to use AI well, the problem has to at least sit in your "known unknowns" and that takes a deliberate discovery mindset. In your team, how does the divide look? 👏 Clap if most people are power users. 👍 Like if it's a 50-50 split. 😂 Laugh if adoption isn't even worth measuring yet.
-
Ryan King liked thisRyan King liked thisI’m thrilled to announce that I’m joining @a16z as a partner to help find and support building the next generation of great consumer companies. I’m excited to join Anish Acharya, David Haber, Alastair (Alex) Rampell, Angela Strange and the entire Apps team. It’s an incredible time to build new AI products for consumers in their daily lives. We're still in the early days but the models are ready, agentic experiences are showing us the way, and consumers are willing to try new things like never before For this era, I’ve been thinking a lot about how I can spend more time helping more companies. I can’t wait to meet teams in the portfolio and in the “future portfolio.” Rather than pitches, I’m most excited to jam on products and growth with teams. I’m eager to help companies navigate the journey from a spark of something working to making something that’s durable. At the same time, this marks the closing of my chapter at Apple. I had the best time helping create products for the world - between working on the App Store and helping start the Games app through leading product marketing and management for Apple Intelligence and Siri. It fulfilled a lifelong dream ever since I had an Apple ][+ as a little boy. Two weeks ago, it was so special to be part of launching the latest version of Apple Intelligence and Siri AI to the world – I can’t wait for you to try it!. I have such gratitude for all of the wonderful people I got to work with, and especially those who invited me to join their teams and projects, and those who joined me and my teams along the way. And now… It’s time to build! Read more here: https://lnkd.in/gmXADpFz
-
Ryan King liked thisRyan King liked thisSoftware is bipedal: code and trust move together, and one without the other just hops along awkwardly. Code works or it doesn't, and you can usually fix it. Trust accumulates slowly and evaporates in an instant, and once it's gone it's hard-to-impossible to get back. XP was a trust factory. Programmer testing, pairing, continuous integration, a customer on the team. Every practice that builds trust also makes you more trustworthy. If I know I'll get paged at night, I do the work so I don't. Single-player, vibe-coded development quietly removes all of those moments. The genie "cares" about satisfying your prompt, not your purpose. What would trust-optimized AI development look like? Slower, on purpose. That's how you actually go faster. Full essay in the comments.
-
Ryan King liked thisRyan King liked thisI’ve just been made aware that there are a few people in the software engineering domain who did not know: - information is not knowledge - knowledge is not experience - experience is not expertise There you go!
-
Ryan King liked thisRyan King liked thisWhy is Meta destroying its engineering organization? Leadership at the social media giant has been on an AI-fueled rampage through its engineering org. We look into what's happened the last two months, all of it self-inflicted: https://lnkd.in/e9KD4vD4
-
Ryan King liked thisRyan King liked this"The Slow" don't exist. They're responding to your incentives. When a project is scoped as "all of all the goals," you've built the perfect breeding ground for slowness, and then blamed the people caught in it. Slash their schedule and they don't get faster; they start hiding progress and inflating estimates. Now nobody knows anything. The missing concept is scope. Not "add a temporal constraint." That just forces lying. Cut scope. How much of which goal, first? Do that and you get sooner, cheaper, and better at the same time, because you've stopped trying to compress the incompressible. And because the landscape keeps shifting, less scope also means more feedback, sooner. Scope is the steering wheel. Most teams grip the schedule and wonder why they can't turn.
-
Ryan King liked thisRyan King liked thisHalf of my system prompt for coding agents is just "please behave like a thoughtful colleague": - Don't guess or assume. Verify against evidence, and ask when it's missing or ambiguous. - Don't affirm ideas reflexively. Stress-test proposals for weaknesses, edge cases, hidden assumptions, and tradeoffs, and push back when warranted. - Offer stronger alternatives when they exist, especially simpler ones. - When surfacing a non-trivial decision, present the question, the options with concrete tradeoffs (use a small snippet or example when it makes the choice easier to grasp), and a recommendation with the reasoning behind it. - Use plain language without sacrificing technical precision. Simplify the prose, not the substance.
Experience & Education
-
Microsoft AI
******* ****** ** ********* *****
-
***** ***
*******
-
*********** *****
*******
-
********** ** *** *********
** ******** *********** undefined
-
-
******** ******* **********
** ******** *******
-
View Ryan’s full experience
See their title, tenure and more.
Welcome back
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
New to LinkedIn? Join now
or
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Publications
Recommendations received
1 person has recommended Ryan
Join now to viewView Ryan’s full profile
-
See who you know in common
-
Get introduced
-
Contact Ryan directly
Other similar profiles
Explore more posts
-
😈 Miloš Rašić
Hash IT • 1K followers
When I was working in online video streaming, we were told we couldn't do CD because TV apps had to be certified by manufacturers. Yet we did CD. It helped us build enough trust so that some manufacturers allowed us to release most changes without certification, and others fast-tracked our certification process. When I was working on internal enterprise software, we were told we couldn't do CD because releases had to be planned and manually tested, deadlines respected, and users trained. Yet again, we did CD. Plans became high-level and abstract as iterative execution took over. Automated acceptance tests and feature flags allowed our QA testers to become our first users in production, rather than gatekeepers for releases. Training was built into the frontend applications. It may be the toughest challenge yet, but I'm determined to make it work in a business SaaS startup, where client demands and approvals are everything. The excuses are plentiful, but I have yet to see a domain where CD has failed to deliver results. After all, some of the most famous CD success stories are about printer firmware and an MMO game. Are you sure the software you are working on is more exotic than that?
19
-
Jonathan Desrosiers
Bluehost • 1K followers
In case you missed it when the episode dropped, I was recently on the Crossword podcast. I finally had a chance to write a bit about our conversations on my site. We covered a lot of ground, including the concept of active versus passive contribution, how to find the appropriate balance between those two groups, the importance of being prepared when new contributors show up, and the nuance between a do-ocracy and a meritocracy. I'd love to hear your thoughts after you listen! https://lnkd.in/evmRqabE
-
Simon Overell
eyeo • 2K followers
I'm feeling a little out of touch so I'm hoping someone in my network can explain this to me... Why do all RAG systems assume Embedding Based Vectors are the best retrieval systems over more traditional / efficient Keyword Based systems like Elastic? A couple of years ago myself and a colleague evaluated a simple RAG system in the finance domain built on LlamaIndex comparing Keyword, Embedding Vectors and Graph based Retrieval. I used RAGAS for the evaluation (which I appreciate is the system marking its own homework) and found Keyword based indexing consistently performed better. My understanding is Vector Embeddings primarily increase recall in the Retrieval stage over Keyword Search. This is only an issue if you have a Recall problem. In many domains basic query expansion seems sufficient to deal with recall. For instance if you have a medium size corpus of a few thousand documents and you want to retrieve all the documents referencing a specific company and interrogate them. Or all documents referencing a small set of say half a dozen companies and compare one aspect of them. In this example keyword search is sufficient and all relevant documents can fit in the LLM's context. Am I missing something about embedding vector retrieval?
36
57 Comments -
Abhay B.
Sprinklr • 2K followers
LLMs are powerful, but they are stateless machines. They don’t carry memory or intent, they work with the context you provide at that moment. The same model can feel brilliant in one scenario and completely off in another. The difference is not the model, but the context and use case we design around it. The technology is already here. The real opportunity lies in how we shape it to solve real problems.
104
11 Comments -
Virendra Singh
1K followers
Jensen Huang just called OpenClaw "the most important release of software, probably ever." When it comes from the man who defined the GPU era, then people listen ( after discounting the vested interest obviously :) While people focus on his quote, we should understand that the real story is adoption speed of Openclaw: it took 3 weeks to reach Linux's 30-year milestone. That's not growth, but an explosion. OpenClaw isn't a tool. It's the operating system for agents that do work. When NVIDIA's CEO calls a project THE thing, every CTO adds it to their roadmap. ByteDance, Alibaba, Tencent are integrating OpenClaw —not for GitHub stars, but because they're terrified of falling behind. On March 6, 2026, nearly 1,000 people queued outside Tencent's Shenzhen headquarters for free, on-the-spot installation of OpenClaw AI. That shows the desperation to be in-step with the technology. In many orgs, OpenClaw is already in production. JPMorgan handles fraud detection with it, Walmart wired it into loan approvals. These are real systems with real money and stakes. The Agenci AI market is fragmented: LangGraph, CrewAI, AutoGen, LangChain, Semantic Kernel, smolagents, PydanticAI. Seven frameworks fighting. But the outcome is obvious. Whoever controls the execution layer controls the decade ahead. The builders who shipped agents in 2024 when LLMs were not ready for it, they already know the significance of these events in 2026 . Which side of this are you on?
73
22 Comments -
Marco Cuturi
Apple • 3K followers
✨✨✨ We have been working with Michal Klein at Apple MLR on pushing a module to train *flow matching* models using JAX. This is shipped as part of our new release of the OTT-JAX toolbox (https://lnkd.in/etJYGjPv) We do so by also introducing the two changes we advocated over the summer to improve FM training, using either (very) large sharp Sinkhorn couplings (https://lnkd.in/eJ6DnF-U) or, even better and efficient, using our semidiscrete coupling approach (https://lnkd.in/e2aBg27a), proposed with Alireza Mousavi-Hosseini and Stephen Zhang. The tutorial to do so is here: https://lnkd.in/e-BAVejK The tutorial doesn't cover yet the conditional FM problem, but the code does, so give it a try, and as always, we welcome your feedback!
205
2 Comments
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content