Articles by Andrew
-
Newsflash: Your CEO Doesn’t Give a Flying Fortran About Velocity
Newsflash: Your CEO Doesn’t Give a Flying Fortran About Velocity
(crossposted on the Jellyfish Blog) It’s time to get honest with ourselves about velocity. Or is it Velocity?.
6 Comments
Activity
12K followers
-
Andrew Lau shared thisWhen Tariq Shaukat and I ran into each other at SFO, we realized we were seeing the same shift from two different angles. AI coding is working. The harder problem now is everything that comes after code generation – verification, software quality, security, and engineering accountability. As CEO of Sonar, Tariq has one of the best vantage points in the industry. Sonar analyzes more than 750 billion lines of code every day across 25,000 organizations, giving him a front-row seat to how AI is changing software development. One of the biggest takeaways for me was this: the challenge is clearly no longer whether AI can write software – it's whether teams can adapt fast enough to keep up. The teams that are getting it right are surrounding it with the right guardrails: – Better context for agents. – Verification built into the development workflow. – Healthier codebases that improve both software quality and future AI performance. We're also asking the wrong questions as an industry. We've spent the past year focused on adoption and token consumption. Those are useful inputs, but they're not the outcome. The outcome is holistic engineering productivity – shipping more value while improving quality, security, and confidence along the way. Watch the full conversation, link in comments.
-
Andrew Lau shared this“Tokenmaxxing” is having a moment – leaderboards, competitions, billions of tokens consumed. Honestly, I get it. Right now, every company is trying to unlock AI impact, and tokenmaxxing and leaderboards focus on the activity to drive AI native patterns - encouraging experimentation, lowering the barrier to entry, and getting teams actually transforming and using the AI native workflows. That matters. But its just the start and is it enough? We also have to make sure we're tokenmaxxing to the right outcome. You can tokenmaxx all day long and still miss the outcomes you actually care about (and spend a lot along the way). The question isn’t just: “How much AI are we using?” alone, it’s: What are we getting from it? Are we shipping faster? Are teams making better decisions? Is the business actually improving? Tokenmaxxing can be a great starting point. It builds the muscle. It drives the direction. It gets people in the game. But the companies that win won’t stop there. They’ll take that usage and turn it into outcomes. More on “tokenmaxxing” via The Wall Street Journal's Isabelle Bousquette, along with insights from Blitzy’s Brian Elliott. Link in comments.
-
Andrew Lau shared thisIn the rapidly evolving world of agentic development, tracking the curve is everything. I recently sat down with Brian Elliott, CEO of Blitzy, to dive deep into the realities of AI transformation. We explored what it actually takes to navigate this shift—and where the industry is headed next.Andrew Lau shared this"The biggest challenge for most organizations in the transformation is no longer just the coding tools." — Andrew Lau, Co-founder & CEO, Jellyfish Our Co-founder & CEO Brian Elliott sat down with Andrew Lau, Co-founder & CEO of Jellyfish, and the conversation landed somewhere most AI discussions never reach. Past the tooling debate and straight into the harder truth: role transformation, change management, and what it actually means to lead an engineering organization right now. Who owns the spec? Where does product management end and architecture begin? And if coding is no longer the bottleneck, what is? Listen to the full episode — link in comments.
-
Andrew Lau shared thisAI is reshaping software development faster than any shift we’ve ever seen . Not just how we write code, but how teams collaborate, how roles evolve, and how organizations think about impact. In my conversation with McKinsey & Company's Martin Harrysson and Prakhar Dixit, I shared a few observations we’re seeing across hundreds of engineering organizations on the Jellyfish platform. Thanks for having me! You can read the full conversation here: https://lnkd.in/eFjWA-bh What we covered: 1. The software development life cycle is being redefined. 2. Adoption drives results. 3. Measurement must be holistic. 4. The next wave of tooling is coming. AI isn’t removing the need for human judgment – it’s increasing it. The companies that succeed will be the ones that evolve their roles, processes, and metrics just as quickly as their tooling.Measuring AI in software development: Interview with Jellyfish CEO Andrew LauMeasuring AI in software development: Interview with Jellyfish CEO Andrew Lau
-
Andrew Lau shared thisEvery engineering leader experimenting with AI eventually runs into the same problem: You can’t improve what you can’t measure. To help, we’re launching the next generation of Jellyfish AI Impact featuring: - Autonomous Agent Insights - Developer Insights - Report Builder - Deeper AI Impact Telemetry in Metrics Explorer This update brings together the full picture across AI-DLC including usage patterns, delivery outcomes, developer experience, and strategic insights so engineering leaders can finally see what’s working, why, and what to do next. Learn more about our newest update to AI Impact here: https://lnkd.in/etSVpyK5New in AI Impact: Autonomous Agent Insights, Developer Insights & Executive Reporting Workflows - JellyfishNew in AI Impact: Autonomous Agent Insights, Developer Insights & Executive Reporting Workflows - Jellyfish
-
Andrew Lau shared thisI’m thrilled to be speaking at GitHub Universe in SF next week! Who’s planning to be there - come and say hi! I’ll be sharing exclusive new data on AI adoption, the impact of AI tools like Copilot on engineering workflows, how to improve gains from AI and more. Then be sure to join me and the Jellyfish team for AfterGlow at Barrio SF. Request an invitation here: https://lnkd.in/ebKsZREY #GitHubUniverse
-
Andrew Lau shared thisAI has reshaped so much of the software development lifecycle that it’s probably more accurate to call it the AI-DLC. Let’s do it. Say it with me. A.I.D.L.C. Let’s see if it sticks! Adoption has accelerated – from 61% of R&D teams last year to 90% today. Teams are moving faster, PR cycles are shorter, reviews are smoother, and engineers are able to dedicate more time to higher-level problem solving and strategy. But the AI landscape is far from settled – and seems to grow more crowded by the day. Most vendors look and feel surprisingly similar, often built on the same models, with differences that come down to usability and user interface. BUT – not all. Some of these matter. And let’s not forget about CONTEXT! Teams are experimenting widely, often running five to nine tools across coding, review, and documentation, which raises both questions of efficiency and cost concerns. The real winners in this space will be the tools that move beyond narrow efficiency gains. They’ll be able to take on larger, more complex engineering tasks autonomously. They’ll integrate seamlessly across the entire SDLC, not just at the point of code generation. They’ll become smarter about leveraging and managing CONTEXT. And they’ll evolve beyond today’s familiar interfaces, offering new experiences that help teams understand and manage the broader impact. I share more about where I think the AI tool landscape is headed in this month’s edition of the Jellyfish Current here 👇
-
Andrew Lau shared thisWe all know AI agents are already showing up in the day-to-day of engineering, but the q is to what extent? To get a clearer picture we looked at code review agents in particular. Jellyfish Research analyzed 1,000 code reviews across 400 companies to find adoption and impact. Here’s what we found – the headlines: Usage: - 93% of teams are trying code review agents today - WOW! - however only 8% of teams have moved past trial mode into regular use Interaction: - agents appear in 22% of all code reviews though a wide distribution – some companies at more than 90% - 18% of agent feedback results in real code change - from a sentiment perspective, 92% of interactions with agents are neutral or positive This is the story of the AI-DLC unfolding in real time – it’s complicated, and nuanced. We’re still early, but the curve is unmistakable – teams are trying these tools and the tools are evolving rapidly. The takeaway: it’s time to experiment – teams should give these tools a try and try again because the tools are evolving. Get a full look at the data here: https://lnkd.in/emFiEBMP
-
Andrew Lau shared this🚨Woot! Jellyfish now with END-to-END visibility across the entire Dev AI SDLC Today we launched three new innovations in AI Impact to give engineering leaders the visibility they need to unlock the full potential of AI in software delivery: - Multi-Tool: Jellyfish unifies visibility on adoption, cost, and impact across multiple AI tools into one consolidated view to identify the most effective tool for different use cases and teams. - Code Review Agent Insights: Jellyfish allows teams to measure the impact of AI code review agents like CodeRabbit, Graphite, and Greptile across the full SDLC. - Dev AI Spend Insights: With Jellyfish’s real-time, usage-based spend data, companies can now tie AI dev spend directly to outcomes to determine ROI at the team, individual, and initiative level. Here’s why this matters: - 90% of engineering teams are already embracing AI tools, but costs are shifting, and we’re all under pressure to make it work – both from a code, product, business value and adoption standpoint. Leaders need ways to disambiguate the noise to understand which tools and agents work for which teams, codebases and projects and why. - As pricing models shift and businesses demand proof of business value, leaders need more – they need real, actionable insight into ROI. Jellyfish AI Impact delivers the insights and visibility engineering leaders need as we move into a fully AI-integrated era in software engineering. See what’s new here: https://lnkd.in/eWvTdfEMJellyfish Launches Industry’s Only End-to-End View of AI’s Impact on Productivity, Quality, and Value Across the SDLC - JellyfishJellyfish Launches Industry’s Only End-to-End View of AI’s Impact on Productivity, Quality, and Value Across the SDLC - Jellyfish
-
Andrew Lau liked thisAndrew Lau liked thisFor many years Boston has been in need of more places where people can build and scale startups. We have always had the talent, what we lacked was the community. Thankfully has truly begun to change. It’s also what struck me when I first met Nathan Spielberg and heard his idea for BUILD617. He and Nicholas Leonard wanted to bring founders into the same room, working side by side, helping each other make things happen. Today, they officially open that room. I’m honored to see it come to life and grateful to have played a very small part. I’m here with Christopher Morrison , who has spent years backing Boston founders and now leads Stifel’s venture banking team in the city, and was thrilled to spend time with Ryan Durkin, who is doing exciting work at WHOOP as well as bringing the Massachusetts AI Coalition to life. They’re the kind of people who make this ecosystem stronger by showing up for it. Nathan and Nick, congratulations. I hope the best measure of BUILD617’s success is how many companies can one day say, “We built it there.” Proud to have been here on day one! HYCU, Inc. #boston
-
Andrew Lau liked thisAndrew Lau liked thisI do not think I could have predicted last night one year ago. A year ago, BUILD617 was just an idea. Last night, I looked around a packed room filled with incredible founders, investors, builders, community members, partners, and friends. To call it surreal is an understatement. I was definitely a little emotional, which caught me by surprise. The energy was undeniable. But really #BUILD617 is not about dinners... or having a physical space or a podcast... Those are just pieces of something much much bigger. This is about building a movement around the people crazy enough to build great things in Boston. And somehow, all of this has happened in less than a year! There are so, so, so many people who deserve credit for making that possible. David Greaney, Kevin Marsh, Rachel Kirby Sandeman, and the entire Synergy team believed early in the vision for a physical home for BUILD617. Dave gets it. He built Synergy here, is a founder himself, and believes deeply in Boston, and he has gone all in on supporting the people building here. We are incredibly lucky to have him in this with us. Christopher Morrison and the team at Stifel Bank have been incredible from the beginning. They are building Stifel's presence here while simultaneously investing their time, network, and energy into BUILD617 and Boston's founder community. Matthew Taylor, MBA, Jill Bosco, Bill Coffill, and the team at CLA (CliftonLarsonAllen) jumped in feet first. They opened their network, showed up, and made it clear from day one that they wanted to help build this alongside us. Eric Popp, Aidan Kittredge, Laura Stoffel, and the team at Cooley LLP have also been huge supporters and have meaningfully helped BUILD617 get to where it is today. And of course, Nicholas Leonard. Thank you for being beside me through all of this. I could not do it without you, and I am incredibly proud of what we are building together. And don't even get me started on the founder wall... There are so many more people I could thank. Honestly, seeing that many people show up last night gave me chills. Way too many people to tag in a single LinkedIn post, but I am going to try to tag a sampling below. Steve Papa Andrew Lau Zenobia Moochhala Chris Combs Chip Hazard Jason Kelly Scott Kirsner Simon Taylor Benjamin Cohen Kairos Shen Matt Thoms Matt Crane Brian Devaney Ryan Durkin Giuseppe Stuto Jonathan Gootenberg Stefanie Chiras, Ph.D. Sara Fraim and so so many more. Tag people in comments! One year ago, BUILD617 was an idea. Today, it feels real. And we are just getting started. Boston, letsgooooooooooooooooooooooooooo 🚀🚀 🚀 🚀 #startups #letsgo #futureisnow #buildthefuture #aistartups #howtocreateamovementinoneyear
-
Andrew Lau liked thisAndrew Lau liked thisHuge congrats to Nathan Spielberg and Nicholas Leonard on the launch of BUILD617's new HQ in Downtown Boston (DTX to be exact). Last night the team held an awesome launch party full of some great builders of the past, present & future here in Boston. I'm very proud of what Nathan & Nick have built. There's a lot more coming from this team. In addition to everything they're doing with their own startups, Tamarin AI and VoiceRun. Come visit us soon! And for more on BUILD617 and ways to get involved check the link in the comments
-
Andrew Lau liked thisHugely exciting milestone!Andrew Lau liked this🎉 We've surpassed 100M PRs analyzed from more than 275,000 software engineers, the industry's largest production engineering dataset. And with insights from more than 90M prompts and 150T tokens to date, Jellyfish data provides the most complete picture of the rapidly changing engineering industry and is regularly cited by industry analysts and journalists. We've also achieved Amazon Web Services (AWS) Generative AI Competency, and we've appointed Julie Neumann as Chief Market Officer to lead the next stage of growth. Get the full story on all of these milestones, including comments from Jellyfish CEO and co-founder Andrew Lau: https://lnkd.in/ebHhFYdn #AIEngineering #AIData #PressReleaseJellyfish Surpasses 100 Million Pull Requests, 90 Million Prompts, and 150 Trillion Tokens Analyzed From Over 275,000 Software Engineers - JellyfishJellyfish Surpasses 100 Million Pull Requests, 90 Million Prompts, and 150 Trillion Tokens Analyzed From Over 275,000 Software Engineers - Jellyfish
-
Andrew Lau liked thisAndrew Lau liked thisThis cake is insane. And I know this is the second cake post in ~ a week... BUILD617 launch tonight. This one is for everyone in the room tonight. Around a year ago this was basically an idea. Tonight we are opening the doors to a real home for Boston founders. The room is going to be packed. The energy is going to be insane. LET'S BUILD BOSTON. 🚀🎂 HUGE thank you to our incredible partners at Stifel Bank Cooley LLP Synergy CLA (CliftonLarsonAllen). More or this soon. #startups #letsgooo #build617 #futureisnow #letsgooo
-
Andrew Lau liked thisAndrew Lau liked thisExcellent talk about agentic software engineering from Andrew Lau of Jellyfish and Jack O'Brien at the Subconscious launch event tonight.
-
Andrew Lau liked thisAndrew Lau liked thisWe challenged our co-op Jeff Krapf to run until a coding agent powered by Subconscious gave up. Jeff ran 9.6 miles during our launch party, and the agent never gave up. Incredible energy last night at the event, and thanks to everyone who came by and make it a huge success. The Subconscious managed inference service is open for business. Special thanks to Andrew Lau for kicking things off with me, and being a champion for Boston startups. The main takeaways I told everyone last night: - Subconscious is an inference API designed to be faster, cheaper, and more accurate for tasks with more than 200k tokens. Now, anyone can grab an API key and get started. - We add the suffix "Marathon" to models on our platform because they're designed for token heavy tasks and keep running without explicit compaction (thanks Jeff for visually demoing). - We've made it *Dead Simple* to use our api with the coding agent you already love. Run "npm install -g subconscious-cli" today. Lot's more coming from Subconscious!
-
Andrew Lau liked thisAndrew Lau liked thisReally enjoyed last night’s Subconscious product launch and the conversation between Jack O'Brien and Andrew Lau about the rise of open models. A few takeaways that stuck with me: open models are quickly closing the gap with closed models and are already good enough for most tasks. As a result, preserving your context window is key to ensuring you can get the most value from the model (which Subconscious can help with!). As AI continues to transform coding, the next challenge will be applying these insights to the rest of the software development lifecycle to unlock even more acceleration. Congratulations to Jack and the entire Subconscious team on the launch, and thanks to the Massachusetts AI Coalition for continuing to bring Boston’s AI community together.
-
Andrew Lau liked thisAndrew Lau liked thisHongyin Luo and I are excited to announce that Subconscious has raised $5.1M to build the inference platform for long-horizon agents. Subconscious was born out of MIT research and frustration that the inference engines are wildly inefficient at tasks that consume lots of tokens over many steps, AKA agents. So, we designed a system specifically for those workloads, and it couldn't have come at a better time. Inference usage is growing 2,500% year over year, agents are consuming the vast majority of those tokens, and open models have nearly closed the gap on closed models. Every megatrend is pointed towards vastly more distributed agents, but they are still too slow, inefficient, and unreliable for many companies and use cases. Not anymore. Using the same chips and the same models, our inference system makes agents 2x faster, 50% cheaper, and up to 10% more accurate on workloads that consume over 200k tokens. These gains are just the tip of the iceberg of what you can optimize when you design an inference system top to bottom with long-horizon agents as the primary workload. If you're using coding agents or building agentic products, Subconscious is another stepwise gain in performance and efficiency. Developers can get started today using our API at https://subconscious.dev AI proliferation is a polarizing topic, but we believe it's going to be extremely positive for the world. Our mission: "allow everyone to do meaningful work and automate the rest without selling your soul". We think reliable, fast, and inexpensive agents will allow people to focus on problems that matter, automate all the monotonous work in between, and then leave plenty of time to enjoy their lives. That's the beautiful world what agents can bring to us, and we deeply believe you shouldn't have to sell your soul to get there. We've been operating for over a year and a half, but today we're finally excited to announce our fundraise, launch our managed inference service, and share that we're growing the team. $5.1M is our total amount raised to date across our pre-seed and seed round. Both rounds were led by MassVentures, with participation from Foothill Ventures, Underscore VC, Oakseed Ventures, Companyon Ventures, Agent Fund, E14 Fund, and Taihill Venture. A special thanks to Stacy Swider who believed in us more than anyone else from day one. Read the full release here: https://lnkd.in/g_GKeip7
Experience & Education
Patents
-
System and method for an affinity capture, user feedback and affinity analysis
Issued US US 9,530,160
-
System and method for manipulating content in a hierarchical data-driven search and navigation system
Issued US 7,325,201
A data-driven, hierarchical information search and navigation system and method enable search and navigation of sets of materials by certain common attributes that characterize the materials. A rules engine provides for manipulation of the content displayed to the user based on the query entered by the user. The rules engine includes one or more rules with a trigger and an action. The action of a rule is performed only if the trigger is satisfied. A trigger may be specified in terms of…
A data-driven, hierarchical information search and navigation system and method enable search and navigation of sets of materials by certain common attributes that characterize the materials. A rules engine provides for manipulation of the content displayed to the user based on the query entered by the user. The rules engine includes one or more rules with a trigger and an action. The action of a rule is performed only if the trigger is satisfied. A trigger may be specified in terms of expressions of attribute-value pairs and is evaluated against a given query or navigation state. The actions can include various techniques for content manipulation, such as supplementing content, rendering content in a particular way, and sorting content in a particular way. An action may be specified in terms of navigation states. The rules engine may include a script for processing the rules.
Other inventorsSee patent
Other similar profiles
Explore more posts
-
Eduard N.
A good next step for AthletAI . At our current stage, the most valuable thing is not confirmation of what we already believe — it’s having our assumptions challenged by people who have seen hundreds of companies grow, fail, adapt, and scale. Looking forward to learning from Igor Ryabenkiy, the AltaIR Capital team, and the other founders in the AltaLab Fall 2026 cohort. Time to get to work.
-
Oliver Alexander
For the first time in US History, it's now possible to search for land based on exactly what you intend to develop. Introducing SearchAI Intentions, by Prophetic. When we started building Prophetic, one of the things that frustrated us most about land search was the invisible prerequisite: before you could find a site suitable for your townhome, hotel, data center, warehouse, etc. project, you had to already intimately know the market. You had to know the zone codes. You had to know which municipalities allowed what uses. You had to translate your vision into zoning language and then go hunting. If you didn't have that local expertise, you were starting from scratch or waiting for a broker to call you. That has been true for every developer, in every asset class, since the beginning of the industry. We just changed how the industry works - nationwide. SearchAI Intentions now lets you search by exactly what you want to build with 120+ development type categories, any market, any jurisdiction, and results in seconds. A hotel developer searches "Hotel - Full Service". A data center developer searches "Data Center". A senior living operator searches "Senior Living - Active Adult" or "Senior Living - Assisted Living". SearchAI reads the actual zoning codes across every zone and jurisdiction in your search area and returns every viable parcel, ranked by fit. It cites the specific municipal code section behind every result so you can act with confidence. What this means practically: for the first time, hotel developers, energy companies, warehouse and logistics operators, data center site selectors, cell tower companies, senior living operators, medical facility developers, commercial and retail developers, economic development agencies, and major brokerages have a land search tool built for how they actually think about acquisition. They can find sites that meet their exact criteria, in any market, in seconds - instead of literal weeks. If you have ever spent days in county GIS systems trying to find viable sites for a project, I would love to show you what this looks like now. For a deeper look at how it works and how Prophetic unlocks development across all verticals, read the full breakdown here: https://lnkd.in/gPzJbt3w #PropTech #LandAcquisition #RealEstateDevelopment #SearchAI #Prophetic #DataCenters #SeniorLiving #EnergyInfrastructure #IndustrialRealEstate #HospitalityDevelopment #homebuilding #cre #realestate #commercialrealestate #realestatebrokerage https://lnkd.in/gz-zrt2P
10 Comments -
Thomas Winter
This week I wrote about intent engineering and frontier operations — how to tell agents what to do, and how to manage the boundary where they work. But there's a layer underneath both that most people skip entirely. Nate B. Jones calls it the harness. Here's the idea: The model is a brain in a jar. It predicts the next token. That's it. Everything else — where the agent works, how it reads your codebase, how it remembers context, how it hands work back to you — that's the harness. And the harness is what actually shapes your working life. When people compare Claude vs Codex, they compare benchmarks. Wrong comparison. The real difference is architectural philosophy: Claude works locally. Incremental handoffs. You see the thinking happen in real time. Codex works in a clean room. Copies your code, thinks privately, slides finished results under the door. Same class of model. Completely different way of fitting into your workflow. This matters because your harness choice is an architectural commitment — not a subscription you cancel next month. It shapes how your team coordinates with AI, how you audit what happened, how you maintain control. The question isn't which harness is better today. It's which harness's evolution trajectory matches where your team is headed. "The model only determines how good your AI is at predicting the next token. The harness determines how usefully it fits into your work." If you're evaluating AI tools and only looking at the model — you're solving the wrong problem. #FieldAI #AIinProduction #BoardToBot #AgenticCommerce #BuildInPublic
-
Pete Jarvis
Check out Roborev: Why? Your Code Review Has a Blind Spot, Roborev Fixes It. Every developer has pushed a commit they immediately regretted. A forgotten debug print. A logic inversion that only looked right at 2:24AM (Chuckle). A change that broke something three files away. Roborev by Wes McKinney (creator of pandas) adds a second pair of eyes to every commit before it ever leaves your machine. That’s its core superpower: pre-push review, automatically, every time. Here (in my view) is what makes it genuinely useful: you don’t need to be using AI coding agents to get value from it. If you write code the old-fashioned way, Roborev still catches things you miss. If you use Claude Code, Copilot, or Codex to generate code at speed. Roborev becomes your spell checker before your output hits your remote repo. That is it’s beauty. The real unlock for teams: this isn’t code review theater. It’s a silent, continuous feedback loop that compounds over time. Fewer regressions. Less context-switching during PR reviews. A virtual reviewer that can flag issues before another human has to peer over your proverbial shoulder. Whether your team writes every line by hand or runs agents at full throttle, the problem Roborev solves is the same: the gap between “code written” and “code verified.” Open source, MIT licensed: https://lnkd.in/gg74dGuZ PS. Wes is also a nice human ;-)
2 Comments
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content