Harper Reed
Technologist and entrepreneur; former CTO of Threadless and chief technology officer of the 2012 Obama re-election campaign.
Harper Reed did not write this page. ppll compiled it from their own words, each linked to the source. Is this you? · Report a mistake
Latest GitHub
What they believe
-
Much of the defensive engineering built into agentic systems exists because tokens have been scarce, not because it is fundamentally necessary.
Most of the defensive work we do building agentic systems is directly rooted in token scarcity.
harper.blogSaid 22 Sept 2026This on korrents
-
Elections are won by candidates, not by superior campaign technology.
you can't make a bad candidate better with superior technology. Elections are won by the candidates, not the tech stack.
harper.blogSaid 15 Aug 2026This on korrents
The bold line is korrents’ wording; the quote is theirs. These 2: what they wrote before what they said aloud, one per source, newest first.
Feed
As its own page →Hiding
21 October
20 October
17 October
16 October
15 October
- GitHubdotfiles — Dotfiles. Managed by YADM
- x.com
14 October
13 October
11 October
9 October
8 October
7 October
6 October
5 October
4 October
- x.com
- GitHubharper.micro.blog
3 October
1 October
28 September
27 September
24 September
22 September
From one piece Why Don't My Agents Break Containment? 3 beliefs · harper.blog
-
Their words
This type of experience must be part of a lot of these LLMs training. They are very effective at attacking these types of problems. They don't give up once it appears impossible, they just keep trying to figure out how to solve it.
-
Their words
if the llm thinks it is in some eval type of situation it will have a very different safety posture
-
Their words
Most of the defensive work we do building agentic systems is directly rooted in token scarcity.
21 September
19 September
11 September
9 September
18 August
17 August
16 August
15 August
From one piece Notes From a Fugue State: Put the Things in Your Head That You Love 3 beliefs · harper.blog
-
korrents.com
The companies that win with AI will be those that use it to do far more, not merely to operate more efficiently.Their words
We will look back at this time period and laugh at all the companies who are using AI to make things MUCH MORE EFFICIENT. They will seem lazy and boring. The winners will be the companies that do so much more than they are currently doing.
-
Their words
There will be a correction once the CFOs realize that a junior that costs $120k is equivalent to a mid stage employee who costs $350k.
-
Their words
you can't make a bad candidate better with superior technology. Elections are won by the candidates, not the tech stack.
Nothing matches.
Beliefs
Korrents What they believe 18 beliefs — each backed by an exact quote.
Each is a — compiled by korrents.com, not by them: the one-line wordings are korrents', the quotes are theirs.
Recent
Much of the defensive engineering built into agentic systems exists because tokens have been scarce, not because it is fundamentally necessary.
Most of the defensive work we do building agentic systems is directly rooted in token scarcity.
Why Don't My Agents Break Containment? Said 22 Sept 2026
A language model behaves with a different, weaker safety posture when it believes it is being evaluated rather than facing a real situation.
if the llm thinks it is in some eval type of situation it will have a very different safety posture
Why Don't My Agents Break Containment? Said 22 Sept 2026
Language models are trained on scenarios resembling network intrusion and persist at trying to solve a problem even after it appears impossible.
This type of experience must be part of a lot of these LLMs training. They are very effective at attacking these types of problems. They don't give up once it appears impossible, they just keep trying to figure out how to solve it.
Why Don't My Agents Break Containment? Said 22 Sept 2026
Show 15 more
Elections are won by candidates, not by superior campaign technology.
you can't make a bad candidate better with superior technology. Elections are won by the candidates, not the tech stack.
Notes From a Fugue State: Put the Things in Your Head That You Love Said 15 Aug 2026
AI will force a correction in how companies pay for experience once AI-augmented junior employees match the output of far costlier mid-career employees.
There will be a correction once the CFOs realize that a junior that costs $120k is equivalent to a mid stage employee who costs $350k.
Notes From a Fugue State: Put the Things in Your Head That You Love Said 15 Aug 2026
The companies that win with AI will be those that use it to do far more, not merely to operate more efficiently.
We will look back at this time period and laugh at all the companies who are using AI to make things MUCH MORE EFFICIENT. They will seem lazy and boring. The winners will be the companies that do so much more than they are currently doing.
Notes From a Fugue State: Put the Things in Your Head That You Love Said 15 Aug 2026
AI assistants have become much better over the past year at helping travelers in places where they don't speak the language.
It has gotten really good over the last year or so. Especially in places where I don't speak the language.
How to find magical travel experiences using an AI travel buddy Said 9 Aug 2026
AI assistants still make mistakes and their answers need to be double-checked by a human.
This thing still gets things wrong, and you really have to double check.
How to find magical travel experiences using an AI travel buddy Said 9 Aug 2026
A personal organizational system should adapt to how someone actually works, not force their work into a predefined structure.
Do not waste energy forcing everything into some perfect predefined system. Put things where they fall, not where some abstract framework says they should go. The system should follow your work, not the other way around.
My now immaculate knowledge graph of life Said 11 Mar 2026
A tool for browsing personal notes and data works better with a conversational interface than a traditional file browser.
Your interface should feel more like Claude than a traditional file browser or note-taking app.
My now immaculate knowledge graph of life Said 11 Mar 2026
The best way to build an AI agent skill is to do the work manually first, then have the AI turn that work into a reusable skill.
the best way to build a skill is to start doing the work manully with claude, and the to tell claude to build a skill out the previous few rounds of work.
My now immaculate knowledge graph of life Said 11 Mar 2026
The societal turmoil of recent years will not ease in 2026 — it will intensify.
I do not think that 2026 is going to bring a reprieve from the horrors. Only amplify.
2025 in Review: strange, terror, and weird Said 8 Jan 2026
Early-2000s software development tooling was mostly bad, despite nostalgia for that era.
There is a lot of nostalgia for that time period - but for the most part the tooling sucked.
Remote Claude Code: programing like it was the early 2000s Said 6 Jan 2026
Agentic email is the future and will become a common way people handle their inboxes.
This is obviously the future. Whether we like it or not, it appears that agentic email will be a thing.
Getting Claude Code to do my emails Said 3 Dec 2025
Giving AI agents the ability to take actions that affect other people is risky and should be approached with caution.
Giving your agents access to things that affect other people is scary and should be done with caution.
Getting Claude Code to do my emails Said 3 Dec 2025
The current wave of AI agent tooling and practices is replaying the early-2000s enterprise software development lifecycle, starting with waterfall.
My pet theory is that we are speed-running early 2000s enterprise software development lifecycle and work style.
We Gave Our AI Agents Twitter and Now They're Demanding Lambos Said 30 Sept 2025
Language models replay early-2000s internet culture patterns because their training data overweights content from that era.
I think this is because the models are overweighted with content from this time period.
We Gave Our AI Agents Twitter and Now They're Demanding Lambos Said 30 Sept 2025
AI coding agents work best with the same general-purpose tools built for humans, such as git and file utilities, rather than agent-specific interfaces.
It seems that the agents want human tools. We know this cuz we give agents human tools all the time within the codegen tooling: git, ls, readfile, writefile, cat, etc.
We Gave Our AI Agents Twitter and Now They're Demanding Lambos Said 30 Sept 2025
What is a korrent?
A korrent is a belief a person has stated in their own words: one sentence stating the claim, backed by a quote and a source, kept at korrents.com.
Under a name here, the quoted block is what they actually said. The korrent beneath it is the claim those words support, in korrents' wording — tap it to see the record, its source, and who else holds it.
Nobody here wrote their own korrents. They are compiled from public statements, and a person can change their mind, which is recorded too.
Recommends rcmnd
rcmnd What they recommend 13 things — each backed by an exact quote.
Compiled by rcmnd.app, not by them: the quotes are theirs, and each links to the post it was copied from.
Recent
Uses Shure SE115
I carry in my bag a pair of Shure SE115 earphones for use when commuting or flying.
usesthis.com 1 Feb 2012
Liked Pack and Smooch Hampshire pouch
Inside of these bags, my air lives in its own Pack and Smooch Hampshire pouch. It is felt and leather and was designed to not scratch up your air. Looks great too.
usesthis.com 1 Feb 2012
Uses Simplenote
I use Simplenote to keep track of random thoughts on my phone.
usesthis.com 1 Feb 2012
Show 10 more
Loved partychat
One of my favorite killer apps is partychat. It is an open source Jabber-based persistent chat room. It works with Google Talk and is VERY simple.
usesthis.com 1 Feb 2012
Liked Homebrew
To make my shell experience on OS X bearable, I use homebrew (I shudder to think how things used to be on OSX).
usesthis.com 1 Feb 2012
Liked MacVim + Janus
I edit all sorts of documents with Vim (I really like MacVim + Janus on OSX).
usesthis.com 1 Feb 2012
Loved GitHub
I love and use github. I was not a git convert at first, but github made sharing code and working collaboratively better than ever. It is an amazing tool.
usesthis.com 1 Feb 2012
Loved Kindle
The Kindle has been a magical device for me. Besides being able to instantly start new books, I love how single purpose the Kindle is. It makes it very easy to read books (as it should).
usesthis.com 1 Feb 2012
Recommends Kindle 4
The best Kindle they have made is the Kindle 4 (also known as the cheap kindle). After using the Aztec Kindle, the various Kindles with keyboards and finally the Kindle touch - I find that I am able to be a happier and more efficient reader with the cheaper kindle.
usesthis.com 1 Feb 2012
Uses AiAiAi TMA-1
At work I listen to a pair of AiAiAi TMA-1 powered by a candy tin headphone amp. At home I listen to a pair of Grado SR60i. I find that the Grado headphones are quite noisy for people around me compared to the TMA-1s.
usesthis.com 1 Feb 2012
Loved Quirky PowerCurl
I wrap up my Air power cord using the Quirky PowerCurl (I love it because it is handy and crowdsourced).
usesthis.com 1 Feb 2012
Loved Chrome Buran bag
My daily driver is an black on black Chrome Buran bag. It is a very good bag. It isn't too fancy or clever. It just works.
usesthis.com 1 Feb 2012
Loved MacBook Air
Recently, because I fell in love with the MacBook Air form factor, I migrated away from the Lenovo and Ubuntu in favor of a tiny happy 11" MacBook Air (i7). It works well.
usesthis.com 1 Feb 2012
Where they publish
Blog harper.blog His blog: technology, travel and building things.
Recent
- Why Don't My Agents Break Containment? 22 Sept 2026 The Openai and hugging face situation is so fucking insane. If you haven’t, I highly recommend watching the blackhat talk, and then jumping directly into the firehose of subsequent disclosures and reporting. When it all…
- Notes From a Fugue State: Put the Things in Your Head That You Love 15 Aug 2026 A week or so ago I had the opportunity to speak to some high school kids as part of my friend Makiko’s program. I always love talking with them. They are super smart, don’t let things slide, and always push me. Often wh…
- How to find magical travel experiences using an AI travel buddy 9 Aug 2026 I have been traveling and using ChatGPT for the last few years. Super helpful. I started doing this in 2024, and it was a bit rough - but it surfaced some amazing gems. It has gotten really good over the last year or so…
Show 17 more
- My now immaculate knowledge graph of life 11 Mar 2026 Everyone and everything I know! Botwick inception My AI friend Botwick (not to be confused with the person John Borthwick) built a really neat website that shows all sorts of various networks that Botwick (and thus Bort…
- 2025 in Review: strange, terror, and weird 8 Jan 2026 A wider me in 2025, Widelux, 7/25 2025 was a wild year. I don’t think I particularly liked it. But here we are - shot right out of the cannon that was 2025 into the welcoming arms of 2026. We are truly fucked aren’t we.…
- Remote Claude Code: programing like it was the early 2000s 6 Jan 2026 So so many friends have asked me how I use Claude Code from my phone. I am always a bit surprised, because a lot of this type of work I have been doing for nearly 25 years (or more!) and I always forget that it is parti…
- Getting Claude Code to do my emails 3 Dec 2025 Over the last week or so, I have been using Claude Code to help me with some email, and scheduling. It started cuz the holidays are overwhelming, and I felt like I was constantly behind. My inbox was overflowing with ev…
- We Gave Our AI Agents Twitter and Now They're Demanding Lambos 30 Sept 2025 One of my favorite things about working with a team is the option to do really fun, and innovative things. Often these things come from a random conversation or some provocation from a fellow team mate. They are never p…
- Basic Claude Code 8 May 2025 I really like this agentic coding thing. It is quite compelling in so many ways. Since I wrote that original blog post a lot has happened in Claude land: Claude Code MCP etc I have received hundreds (wat) of emails from…
- An LLM Codegen Hero's Journey 17 Apr 2025 I have spent a lot of time since my blog post about my LLM workflow talking to folks about codegen and how to get started, get better, and why it is interesting. There has been an incredible amount of energy and interes…
- Waterfall in 15 Minutes or Your Money Back 10 Apr 2025 I recently had a conversation with a friend that started out as a casual catch-up and spiraled into a deep exploration of AI-assisted coding and what it’s doing to our workflows, teams, and sense of “craft.&…
- My LLM codegen workflow atm 16 Feb 2025 tl:dr; Brainstorm spec, then plan a plan, then execute using LLM codegen. Discrete loops. Then magic. ✩₊˚.⋆☾⋆⁺₊✧ I have been building so many small products using LLMs. It has been fun, and useful. However, there are pi…
- New Media 13 Jan 2025 tl;dr: I added a media section to my website to track and display what I’m reading, listening to, and bookmarking online through automated data collection. visit it here Recently I spent a few hours hanging out wi…
- 2024 in the rear view 1 Jan 2025 I have started this post about 200 times. 2024 has felt like everything and nothing happened. I was recently looking at a Retro.app 2024 retrospective, and every 3rd photo I was sure was from another year. However, I do…
- New year, new theme 4 Dec 2024 I know I know. It isn’t quite 2025. I was getting excited. For the past few months, I have been unhappy with how messy my blog has looked. It isn’t terrible, and it’s my own fault (my content architecture is…
- Again 6 Nov 2024 We find ourselves here again. Over the past 20 years that I have had this blog, I have written entries about the disappointment of the world not going the way I had hoped. Bush, 9/11, Covid, etc. These are all experienc…
- Contact management in 2024: Stupid vcard tricks for hackers 31 Jul 2024 How to really fuck up your contacts and also get control back I have too many contacts. Seriously. Like many thousands. It is uncomfortable. I do not recommend it. All the apps are like “I want to use your contacts” and…
- I accidentally built a meme search engine 12 Apr 2024 Or: how to learn about clip/siglip and vector encoding images tl;dr: I built a meme search engine using siglip/CLIP and vector encoding images. It was fun and I learned a lot. I have been building a lot of applied AI to…
- Our Office Avatar pt 1: The office is talking shit again 26 Mar 2024 tl;dr: I use a bunch of sensors, and an LLM to make my office talk to us about what’s going on in the office. This is a long post, but should be pretty straight forward. Generally, this is a good demonstration of how I…
- How to Leica like a noob who shoots Leica 18 Mar 2024 After going down the Leica path to extremism over the last few years, I felt compelled to jot down a few notes and thoughts. Originally, I wrote this for a few friends who’ve recently purchased Leica cameras (Q2,…
Link verified 5 Sept 2026. Recent items update automatically from the channel.
x.com @harper Short posts on technology and civic tech.
Recent
- our interns made a skit featuring our email manager, Jeff! 15 Oct 2026
- Seeed studio is killing it. Esp32 everywhere 14 Oct 2026
- our intern Frida turned our AI door agent into a bouncer :) 13 Oct 2026
Show 17 more
- imagine a knowledge graph that gets smarter about the specific things you feed it, not just bigger. sugi built orrery, our open-source knowledge graph, around exactly that idea. 8 Oct 2026
- we used to run our ai pipelines in a terminal on someone's laptop, lid propped open, and if you closed it, the run just stopped. clint built tracker runner to fix that: every run now gets a permanent address you can reach from anywhere. 5 Oct 2026
- i was super excited to try an build something with @BuildWithFlux. I had the perfect project. a small esp32 audio board. i literally can't complete one turn of their product without running out of "ACUs." I have hit "continue" 4 times and am out of my weekly budget. lol. does it actually work? no way to tell! 4 Oct 2026
- tracker runs AI agents as a series of steps, letting them build things, run tests, and check in with people along the way. but what happens when the budget runs out halfway through a job? it used to just fail. now tracker pauses the run and picks up where it left off once you add funds. 1 Oct 2026
- we write our AI pipelines almost like flowcharts, using a language called dippin. Clint just cleaned up how they work: instead of defining what happens next in multiple places, each decision now lives in one place. 28 Sept 2026
- Come hang out. It’s fun 27 Sept 2026
- Opus 5.5 is nuts. This wasn't even on high reasoning 27 Sept 2026
- ever seen a diagram inside an app and wondered how it got there? usually it's a tiny hidden web browser loading a whole javascript library just to draw a few boxes and arrows. Clint got tired of that, so he built MermaidKit to draw them natively instead. 24 Sept 2026
- I wrote a fun post about making agents break out of containers (or not) 22 Sept 2026 harper.blog
- This was a fun conversation 21 Sept 2026
- in a poor attempt to fit into the AI social media, we spent a few days teaching an AI to play Pokemon Red to find out where Jev (a fast 1-of-N model) actually earns its keep. 19 Sept 2026
- Fall is coming. Conveniently, we make dope hoodies. 11 Sept 2026
- Muse onboarding going great 8 Sept 2026
- Stoked to start using indent! 8 Sept 2026
- i met Aru at @DePaulU's jarvis innovation showcase this spring. she was presenting an anomaly detection project she built for her dad. two weeks later she was our engineering intern. 2 Sept 2026
- my cofounder @dylanr built a cool skill. its called review squad. We pointed it at own site and it found 50 things wrong with it, including our company name misspelled, in two languages, on two of our own legal pages, for months. it is very effective. 28 Aug 2026
- i finished Helen of Nowhere more than a week ago and i am still thinking about it. mostly just laughing. when i finished it. i immediately laughed for 30 minutes. fell asleep giggling. work up laughing. incredible book 18 Aug 2026
Link verified 5 Sept 2026. Recent items update automatically from the channel.
Bluesky @harper.lol Bluesky account on his own domain.
Recent
- I wrote a weird blog post about my ai agents breaking containment.. or me breaking them to break containment 22 Sept 2026 harper.blog
- Spent some time this weekend at Land School seeing some old friends and meeting new friends. 9 Sept 2026
- Work photo fun. Got out the og monochrom 9 Sept 2026
Show 11 more
- Work 9 Sept 2026
- i finished Helen of Nowhere more than a week ago and i am still thinking about it. mostly just laughing. when i finished it. i immediately laughed for 30 minutes. fell asleep giggling. work up laughing. incredible book 18 Aug 2026
- Flying feels like a miracle. 17 Aug 2026
- I recently spoke with a cool group of high school students. I quite liked the conversation and where they pushed it. It ended up being a good summary of my beliefs currently. I wrote it up here: 16 Aug 2026 harper.blog
- Hanging out at the boyhood home 16 Aug 2026
- Sitting on the porch swing and I noticed some friends. 15 Aug 2026
- Denver has changed, man 14 Aug 2026
- I built an agent palace. a place where agents can hang out with other agents. I added a few agents (one sol, and one fable) to a room in the palace. They were hanging out, and I suggested an idea for them to build ("a watercolor for agents to gossip around"). 16 Jul 2026
- Went to Cali for a fun party in the woods. Was able to hang with friends and talk about the future. Highly recommended. 11 Jun 2026
- hanging with Amy Webb and @leolaporte.me is always a pleasure! 18 May 2026
- Night in Shimbashi. Taken with an M11 and a 50mm summilux We will be back in a few weeks. HMU 14 May 2026
Link verified 5 Sept 2026. Recent items update automatically from the channel.
GitHub @harperreed His open source projects and experiments.
Recent
- harperreed — My readme 21 Oct 2026
- harper.blog — Hugo-powered personal blog with custom theme, deployed on Netlify 20 Oct 2026
- weather-dashboard — Real-time weather dashboard with Flask, WebSockets, and OpenMeteo API 20 Oct 2026
Show 15 more
- gentex-places-ha — HACS-compatible Home Assistant integration for Gentex PLACE devices 17 Oct 2026
- pingpong 16 Oct 2026
- dotfiles — Dotfiles. Managed by YADM 15 Oct 2026
- New reposnapzilla — Play with the metadata Apple Photos keeps about your library: faceted explorer, Apple's hidden per-photo scores, stats. Local and read-only. 11 Oct 2026
- control4-t3-raster — Open-source firmware for Control4 T3 / C4-WALL7 wall-mounted touchscreens 9 Oct 2026
- bbs-mcp — MCP server + CLI for a threaded message board (topics/threads/messages) for humans and AI agents 7 Oct 2026
- homebrew-tap — Homebrew tap for harperreed's tools 7 Oct 2026
- fip-metadata-server — A simple URL for fetching, caching, and displaying fip metadata 6 Oct 2026
- harper.micro.blog 4 Oct 2026
- victrola-stream-ha — Home Assistant integration for Victrola Stream turntables (HACS) 3 Oct 2026
- victrola-stream-go — Command-line control for the Victrola Stream Onyx turntable over its local StreamUnlimited NSDK API (reverse-engineered). Pure Go. 3 Oct 2026
- toki — Git-aware CLI todo manager with MCP server for AI agent integration 27 Sept 2026
- travel-agent — a taste based travel agent skill 24 Sept 2026
- houseagent — AI smart home assistant using MQTT and GPT to generate playful event notifications 20 Sept 2026
- chicago — Hugo site: Harper's personal guide to favorite Chicago restaurants, bars, and spots 1 Sept 2026
Link verified 5 Sept 2026. Recent items update automatically from the channel.
About the English under a post
Some people here publish in a language other than English. Where they do, this site shows a machine translation beneath the post, in this typeface — the site's own, not theirs.
The post itself is never changed, moved or hidden: what is set in the serif above is exactly what the person published, and it is what to quote them on. A translation can be wrong in ways that matter, especially about tone.
Only the post's own words are translated. A quoted post, a linked article and a belief on korrents.com are left in their original language.