Articles by Thariq
-
Lessons from Building Claude Code: How We Use Skills
Lessons from Building Claude Code: How We Use Skills
Skills have become one of the most used extension points in Claude Code. They’re flexible, easy to make, and simple to…
21 Comments
Activity
17K followers
-
Thariq Shihipar reposted thisThariq Shihipar reposted thisHow do you actually build an effective harness with Claude? We had Thariq Shihipar from Anthropic at Navigate 2026 to talk about "Unhobbling Claude", the difficulties and processes on how to build agents and harnesses. Watch the full talk at the link below.
-
Thariq Shihipar reposted thisThariq Shihipar reposted thisHow do you actually build an effective harness with Claude? We had Thariq Shihipar from Anthropic at Navigate 2026 to talk about "Unhobbling Claude", the difficulties and processes on how to build agents and harnesses. Watch the full talk at the link below.
-
Thariq Shihipar reposted thisThariq Shihipar reposted thisI’m hiring Anthropic’s first Head of Technical Training to build our delivery function, and a Lead Technical Instructor to teach alongside them. Links to both roles are in the comments! The goal for both is to make instructor-led training on Claude as good as the one pictured. At Gagan Bhat’s Code with Claude workshop, a room of engineers built their first Claude Managed Agent in under an hour. What made this particular workshop great: 1. A real problem. It opens with a 2am page for a checkout latency spike. Attendees build the SRE agent that finds the cause. 2. Approachable. A repo handled setup and shipped a working dashboard, so people only wrote the agent: seven functions, about 38 lines of code. 3. Complete. It covered every core Managed Agents concept, in the order the docs introduce them. 4. Fun to build. The agent came online one function at a time, then searched a 70,000-line log and named the commit that caused the outage. 5. A way back for anyone stuck. A finished reference file let people catch up without stopping the room. 6. Something to keep. Swap the mock data for real monitoring tools and the same code works. 7. Tested and rehearsed. An end-to-end test proved the lab worked before anyone ran it. Gagan had practiced the delivery and kept the room engaged throughout. I need two people who can deliver curricula like this, and teach others to do it too. Head of Technical Training, San Francisco You build the function. You spend your first year mostly teaching, from 12-person labs to 100-person field-readiness days. You’ll hire and manage an instructor bench, define what “ready to teach” means, and decide who is cleared to teach, both inside Anthropic and through our training partners. Lead Technical Instructor, New York or San Francisco Half your time is hands-on teaching, in person and remotely. The other half is making other instructors better through observation, coaching, and train-the-trainer programs. You report to the Head of Technical Training. Both roles need an engineer who teaches. You’ve shipped production software. You’re familiar with LLM APIs, agents, tool use, and MCP. You can build an agent in front of a skeptical staff engineer and fix a broken lab while the room waits. Who’s the best engineer you know who also teaches well? If it’s you, DM me a link to something you taught. Links to both roles are in the comments.
-
Thariq Shihipar reposted thisThariq Shihipar reposted thisWe are month out from Navigate 2026, Browserbase's conference for the best teams shipping agents. I'm really excited to announce some of the incredible speakers we have lined up: > Malte Ubl, CTO @ Vercel > Thariq Shihipar, MTS @ Anthropic > Monica Bajaj, SVP of Engineering @ Okta > Ian Livingstone, CEO @ Keycard On stage in SF on September 10th. We're almost at capacity.
-
Thariq Shihipar reposted thisThariq Shihipar reposted thisWe are month out from Navigate 2026, Browserbase's conference for the best teams shipping agents. I'm really excited to announce some of the incredible speakers we have lined up: > Malte Ubl, CTO @ Vercel > Thariq Shihipar, MTS @ Anthropic > Monica Bajaj, SVP of Engineering @ Okta > Ian Livingstone, CEO @ Keycard On stage in SF on September 10th. We're almost at capacity.
-
Thariq Shihipar reposted thisThariq Shihipar reposted this“We’ve got /loop, /goal, and workflows. They all get the agent to run for long periods of time.” Here’s my new episode with Thariq (Anthropic) where he shared how he plans, builds, and runs loops with Claude Code: → How /loop, /goal, and workflows work differently → Live demo: Editing launch videos with Claude → Using HTML artifacts to plan and learn Some quotes from Thariq: “We cut Claude Code’s system prompt by 80%. As models have gotten smarter, they need less direction, fewer constraints, and fewer examples.” “Planning is an iterative process of exploring, investigating, and finding out what you don’t know.” “One failure mode I see is that people glaze over AI’s plans. You want to make sure it’s something you read.” I've been wanting to interview Thariq for a long time and I learned so much from this epsiode. 📌 Watch now: https://lnkd.in/gw93_g2A Thanks to our sponsors: Riverside: All-in-one AI studio for podcasts and video https://lnkd.in/gY3rxbG2 Wispr Flow: 4x faster than typing with your voice https://lnkd.in/gsv8Vk6K Also available on: Spotify: https://lnkd.in/gSJXspzG Apple: https://lnkd.in/gqq6M38r Newsletter: https://lnkd.in/gf8m8MRC
-
Thariq Shihipar shared thisMy keynote at AI Engineer World Fair: “A Field Guide to Fable” is live on YouTube! I talk about how working with Claude changes with Fable and how important it is be more ambitious. https://lnkd.in/gY5w-8HV
-
Thariq Shihipar shared thisFor the Claude Blog I wrote about our new feature: dynamic workflows and how it enables you to do a whole new set of tasks in Claude Code. Here are some example prompts from the post that might get you thinking: - "Take my business plan and run a workflow where different agents tear it apart from an investor's, a customer's, and a competitor's perspective." - "Here's a folder of 80 resumes, use a workflow to rank them for the backend role and double-check the top ten. Interview me using the AskUserQuestion tool for a rubric." - "Using a workflow, go through my last 50 sessions and mine them for corrections I keep making and turn the recurring ones into CLAUDE.md rules" - "I need a name for this CLI tool. Use a workflow to brainstorm a bunch of options and run a tournament to pick the top 3." - "Go through my blog post draft and using a workflow verify every technical claim against the codebase, I don't want to ship anything wrong." https://lnkd.in/gHfFzHkdA harness for every task: dynamic workflows in Claude Code | Claude by AnthropicA harness for every task: dynamic workflows in Claude Code | Claude by Anthropic
-
Thariq Shihipar reposted thisThariq Shihipar reposted thisThariq Shihipar is a builder's builder. After research at MIT, exiting his company, raising millions for another, and exploring research threads as an SPC member, he's now on the team building Claude Code. He's coming back to South Park Commons next week.
-
Thariq Shihipar reacted on thisThariq Shihipar reacted on thisLife update: dropping the "Visiting" — I'm joining Y Combinator as a General Partner 🤩 YC was my first choice when I applied with Cron in 2019. Seven years later, I'm choosing it again. This past year as a Visiting Partner, I got to come full circle: working with founders through the W26 and S26 batches, helping them turn early ideas into products people love. I'm incredibly excited to keep finding founders and helping them build the iconic companies of this generation 🔥
-
Thariq Shihipar reacted on thisThariq Shihipar reacted on thisHow do you actually build an effective harness with Claude? We had Thariq Shihipar from Anthropic at Navigate 2026 to talk about "Unhobbling Claude", the difficulties and processes on how to build agents and harnesses. Watch the full talk at the link below.
-
Thariq Shihipar reacted on thisThariq Shihipar reacted on thisThere is no trick to building for agents. I get asked constantly: "How do we write better docs for agents?" My answer is always the same: just write better docs. This week I got to speak on this topic on stage at the WeAreDevelopers conference. A few of the things I said on stage: 1) Building for agents mostly means building for humans. Great AX is usually just a great product people love, with tools exposed for agents. That's it. 2) Behind every agent is a person trying to get something done. It's easy to forget that and optimize for the agent instead of the problem the person behind it actually has. 3) UI will matter less in a world of agents. But "UI is obsolete" is far overblown. No one wants a world where it's only chatboxes. 4) You still want to read a good book once in a while. Nobody wants to live inside a chat box. 5) Across the content we serve, most of the traffic already isn't human. 67% of it is agents, and it's heading toward 90%. That changes how we build, not who we build for. It's really important to remember that while the agent is the executor. The human is still the decision maker
-
Thariq Shihipar reacted on thisThariq Shihipar reacted on thisBefore OpenAI and Jev, I took a year off to travel, meditate, and play League of Legends. That year shaped TypeSafe AI's culture. It sounds ridiculous, but after I left Google Brain in 2018, I took a year and a half off to figure out what actually made me happy. I threw myself into anything and everything. I traveled the world, dabbled in improv, and did a lot of art and meditation. I even sweated my way to top-30 in League of Legends and top-2 in Hearthstone (maybe even harder than building Jev). But none of it made me feel happier or truly fulfilled. The dopamine hits from pentakills only get you so far… What I really missed was working on sick-ass projects with driven people. So I came back to SF and joined OpenAI in 2020, when it was still a small obscure research lab. It reignited my love for AI. We were a bunch of nerds genuinely driving the frontier of AI forward. I worked with great people on RLHF, InstructGPT and ChatGPT. It was so cool! I learned there’s nothing cooler than working on the most exciting stuff with people just as engaged as you are. So when I started TypeSafe, I built our culture on 2 principles: Doing exciting work at the frontier of AI Building a team of smart, driven people who push each other to be better So far, we’ve delivered on both. Jev’s launch been absolutely insane and we have so much more coming up soon 👀 If that sounds like fun, we’re hiring!
-
Thariq Shihipar reacted on thisThariq Shihipar reacted on thisAI lets you try more ideas than you'll ever ship. Each one can teach you something. Combine what you learn and that's what ships. The mistake is treating cheaper code as a reason to ship more of it. Most of what you build now should be throwaway, right? A prototype that answers "does this feel right?" in an afternoon. Or maybe an attempt that proves an approach won't scale before you've committed to it. Three versions of a flow put in front of a handful of users so you can watch where they get stuck. None of that belongs in prod. It might belong in your head, team discussion or A/B tests. The output of all this exploration isn't features. It's judgment. You end up knowing why the obvious approach fails, which edge cases your users actually hit and what they reach for when nobody's explaining things. That understanding is what makes the thing you finally ship better. So the leverage isn't 10x the features. It's 10x the learning per feature. Ship less, understand more, and make the thing you do ship count. #ai #programming #softwareengineering
-
Thariq Shihipar reacted on thisThariq Shihipar reacted on thisExactly. Excellent take from Thariq Shihipar AI and coding agents don't give you a license to become a feature factory. Users are paying you for convenience, not for the cognitive load of digging through a hundred features to find what they need.
-
Thariq Shihipar reacted on thisThariq Shihipar reacted on thisExcited to share that Ando's officially out of stealth & we’ve raised $20M from Accel, Index Ventures, and Emergence Capital to build a new, AI-native team messaging platform (Slack / Teams built from the ground up for the AI era). - TechCrunch: https://lnkd.in/g3jzqZnD - Accel's blog: https://lnkd.in/gnGUeGAM - Index's: https://lnkd.in/gntxEDmh I started my first company, Alloy, when I was 19, and spent five years running it as CEO. Half a decade is a meaningful amount of time to spend building one company, and you inevitably learn a lot about your own founder-market fit - what kinds of markets reward your particular strengths, what problems you’re willing to think about for years, and what kind of work gives you energy in the morning. I realized I’m happiest when I'm working at the frontier of tech and hammering away on a product I personally use. I left Alloy amicably, with no concrete plans besides reading papers and tinkering. Around the same time, Anthropic released MCP. It felt like a missing piece: LLMs could finally use tools in a general enough way to actually do work. I got pretty obsessed with the idea and spent the beginning of 2025 building early agents and MCP servers for companies. Pretty much everyone wanted those agents in Slack. So I spent a lot of time trying to make them work, but kept running into the same problems. Initially I thought they were mostly technical constraints I could work around - inefficient context through the API, permissions, etc. But the deeper I went, the more it seemed like the underlying architecture itself was wrong. We were taking something that could behave like a coworker and treating them as apps and bots you had to “install.” Eventually, it seemed easier to rethink the messaging platform than to keep working around those assumptions. That became Ando. In Ando, agents are participants in the team. They have their own identity, onboarding, permissions, and product experience. They can be present across conversations, understand what’s happening, and take on work alongside the rest of the team. Today, teams across 15 countries are using Ando. We’re still early and figuring out a lot, but it’s been fun seeing how quickly some teams are changing the way they work around agents. For now, we’re focused on onboarding 2–40 person teams that are actively using cloud agents and want them working alongside the rest of the team. A few asks: 1. Use Ando. If your team is small & already using cloud agents heavily, I’d love for you to try it. Email me and I’ll get you on. If not, make sure you sign up to the waitlist 2. Come work with us. We’re still a very small team with a lot to build: https://lnkd.in/diDRimbT 3. Tag people. If a team or person comes to mind who should know about Ando, please send this their way 🙏🏻
Experience & Education
Other similar profiles
Explore more posts
-
Mike Duboe
Greylock is leading the seed in SuperMe, the AI-native professional network. Most expertise is not yet legible to LLMs. Professionals demonstrate expertise in private contexts: product docs, meeting notes, and most often - in their own heads. SuperMe is building a way to extract and structure these insights into AI profiles, grounded in an expert's real words and personality. And it is network-first; the more conversations one's AI profile has, the more opportunities for experts to reinforce or create new insights -- expertise compounds. This is one of the strongest examples of founder-market fit we've seen at Greylock. Casey Winters is the most in-demand product & growth advisor that marketplace operators (self included) wish they could hire. And Ludo Antonov built & led engineering teams at some of the most iconic network businesses over the past decade: Whatnot, Lyft, and Pinterest (where he worked closely with Casey). The two started this company out of our offices, and have until now been quietly building a highly curated network with profiles like Elena, Lenny, Brian, etc. We see a unique opportunity to create the professional network graph in a world where likes & follower counts have lost meaning. More on our investment below:
13 Comments -
Chris Talley
The most recent Queued Up report from Berkeley Lab is out. Steven Zhang and I are proud to be listed as coauthors on this edition. The Interconnection.fyi team worked closely with Joseph Rand and the rest of the LBNL team to provide the data that powers this year's analysis. Supporting this level of research is core to our mission of bringing transparency to the wholesale energy markets. The 2025 edition (covering data through the end of 2024) highlights some significant shifts in the landscape: Active Capacity: 2024 closed with nearly 2,300 GW of generation and storage seeking interconnection. Changing Mix: Active natural gas capacity increased by 72% year over year, while solar and storage saw slight decreases in total queue volume. The Backlog: 408 GW of capacity already has an executed or draft interconnection agreement but has not yet reached commercial operations. Timelines: For projects built between 2018 and 2024, the median duration from request to operation has doubled compared to the early 2000s. This report is the definitive annual benchmark for the industry and provides a vital baseline as we begin to see the implementation of FERC Order 2023. The analysis in this report is based on our EOY 2024 data snapshot. Since then, we have continued to track and update these queues every single day. If you want a live view of how these numbers have shifted in the months since this snapshot was taken, follow Interconnection.fyi and subscribe to our Substack. You can find the full slide deck, interactive maps, and raw data files at the link in the comments.
3 Comments
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content