{"id":39270,"date":"2026-09-13T21:04:40","date_gmt":"2026-09-13T21:04:40","guid":{"rendered":"https:\/\/www.duck9.com\/blog\/?p=39270"},"modified":"2026-09-13T17:04:45","modified_gmt":"2026-09-13T21:04:45","slug":"agent-harnesses-are-the-new-gpt-wrappers","status":"publish","type":"post","link":"https:\/\/www.duck9.com\/blog\/agent-harnesses-are-the-new-gpt-wrappers\/","title":{"rendered":"Agent Harnesses Are The New GPT Wrappers"},"content":{"rendered":"
\n
\n
\n
\n
 by Greg Isinberg<\/div>\n
<\/div>\n
AGENT HARNESSES ARE THE NEW GPT WRAPPERS<\/div>\n
<\/div>\n
What exactly is an agent harness? A harness does 4 things:<\/div>\n
<\/div>\n
1. Runs the model in a loop so it keeps working step after step instead of answering once and stopping.<\/div>\n
<\/div>\n
2. Gives it hands to read files, call tools, open portals, and run code.<\/div>\n
<\/div>\n
3. Manages its memory so hour three of a job still knows what happened on hour one.<\/div>\n
<\/div>\n
4. Enforces the rules about what it can touch and when it has to stop and ask a human.<\/div>\n
<\/div>\n
WHY IT’S INTERESTING<\/div>\n
<\/div>\n
– The market is MEGA. I think about it like a wrapper let you sell software ($800B market) but a harness lets you sell the work ($5T+ market)!<\/div>\n
<\/div>\n
– It doesn’t depend on any one company’s model. The knowledge about the job lives in the harness, so you can run GPT today, Claude next month, an open model like Google Gemma, Qwen, Deepseek etc on your own machine after that, and it keeps working. A wrapper was one model doing everything and a harness is a router.<\/div>\n
<\/div>\n
– It gets better the more you use it. See, every correction a human makes becomes a rule the harness keeps, so what you have after 500 jobs is a completely different product than what you had after 5. Wrappers only got better when OpenAI got better.<\/div>\n
<\/div>\n
– You can charge for finished work! The harness knows when a job is done, so you can price per claim, per filing, per review, per resolved exception, or per closed month. This helps compete with SaaS!<\/div>\n
<\/div>\n
– Most jobs are just the same handful of decisions, repeated, using the same few tools. That’s true for most of the 800+ occupations out there, which is why almost all of them could have a harness built for them. Lots of space for startups to be building.<\/div>\n
<\/div>\n
-The frontier labs won’t come for these. OpenAI is not going to learn how a freight claim gets denied in Rotterdam or which prior auth your specific payer rejects. Those markets are too small for them and the knowledge only exists inside the job. They’ll keep making the models better, which just makes your harness better which is cool.<\/div>\n
<\/div>\n
– OH, AND really important, you can ACTUALLY build one now. OpenAI’s Agents API, Claude’s managed agents, Vercel’s eve, LangChain harness framework etc. They all make it way easier to build agent harnesses. <\/div>\n
<\/div>\n
Agent harnesses truly are the new GPT wrappers.<\/div>\n
<\/div>\n
This is just the beginning.<\/div>\n<\/div>\n
<\/div>\n
\n
<\/p>\n\n\n\n\n\n
\"HSHcd8iXkAQGOq3.png\"<\/td>\n<\/tr>\n
\n
AGENT HARNESSES ARE THE NEW GPT WRAPPERS What exactly is an agent harness? A harness does 4 things: 1. Runs the model in a loop so it keeps working step after step instead of answering once and stopping. 2. Gives it hands to read files, call tools, open portals, and run code. 3. Manages its mem\u2026<\/div>\n<\/td>\n<\/tr>\n
\n\n\n\n
<\/a><\/td>\n\n
\n
GREG ISENBERG (@gregisenberg) 813 likes \u00b7 63 replies<\/font><\/a><\/div>\n
x.com<\/font><\/a><\/div>\n<\/div>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

<\/a><\/div>\n<\/div>\n

<\/div>\n


<\/p>\n

\n
Chapter 1 to Chapter 14\u2019s an \u201cEaster Egg\u201d at <\/span>#ch1 to #ch14. Including #ch2 which\u2019s chapter 2 at my house in Napa California<\/span><\/div>\n

http:\/\/www.youtube.com\/watch?v=ejeIz4EhoJ0<\/span><\/p>\n


<\/span><\/div>\n
On 09-09-39, “What They Will NEVER Teach You at Stanford Business School” debuts at 300 w 44th St at New York Fashion Week’s front row
http:\/\/www.youtube.com\/watch?v=QXIaNZi3mHQ<\/span><\/span><\/div>\n

What A Super Model Can Teach a Harvard MBA About Credit www.slideshare.net\/larrychiang\/what-a-super-model-can-teach-a-harvard-mba-about-credit<\/a><\/span><\/p>\n

American Express hosts me mentoring you about FICO scores at New York Fashion Week
t.co\/inxTmZAj<\/a><\/span><\/p>\n

My video boils down 20,000 hours and moves you to the right on the entrepreneur bell curve 
http:\/\/www.youtube.com\/watch?v=eudADPfTWiE<\/span>
<\/span>***********<\/span><\/div>\n

Steve Jobs Texted me on 650-283-8008 in the same way that Mr Jobs called Bill Hewlett <\/span>https:\/\/x.com\/superSaiyanSkai\/status\/1941392367304761636\/video\/1<\/a><\/span><\/div>\n

<\/p>\n


<\/span><\/div>\n
Larry Chiang<\/span><\/div>\n
Fund of Founders<\/div>\n
Founding Stanford EIR<\/div>\n
@duck9 alum, Deeply Understood Capital Credit Chinese Knowledge 9<\/div>\n
Solo Founder Uber API<\/div>\n
650-566-9600 Office<\/div>\n
650-566-9696 Direct<\/div>\n
Cell: 415-720-8500 <\/div>\n

<\/span><\/p>\n

650-283-8008 (cell)<\/span><\/div>\n

<\/span><\/div>\n
Editor of the widely syndicated “What They Don’t Teach at School”
whattheydontteachyouatstanfordbusinessschool.com\/blog<\/a><\/span><\/p>\n

CNN Video Channel: ireport.cnn.com\/people\/larrychiang<\/a><\/span><\/p>\n

Read my last 10 X posts at www.X.com\/LarryChiang<\/a><\/span><\/p>\n

Author of #WTDTYASBS a NY Times Bestseller released 09-09-09 at #NYFW on a runway under the tents
whattheydontteachyouatstanfordbusinessschool.com\/blog\/?s=Ny+times+bestseller<\/a><\/span><\/p>\n

www.fastcompany.com\/embed\/c0d4562ea2049<\/a><\/span><\/p>\n

52 Cards. Two Jokers. What They DO Teach You at Stanford Engineering
http:\/\/www.youtube.com\/watch?v=vDBY0GkI3-g<\/span><\/p>\n

Emergency swings and cutting deals as an 9 year old
http:\/\/www.youtube.com\/watch?v=OFGY7v9C4G0<\/span><\/p>\n

Hunter Pence shared thoughts before winning WORLD SERIES’ Game #7
http:\/\/www.youtube.com\/watch?v=usu0luYy9pw<\/span><\/span><\/p>\n


<\/span><\/span><\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n","protected":false},"excerpt":{"rendered":"

 by Greg Isinberg AGENT HARNESSES ARE THE NEW GPT WRAPPERS What exactly is an agent harness? A harness does 4 things: 1. Runs the model in a loop so it keeps working step after step instead of answering once and stopping. 2. Gives it hands to read files, call tools, open portals, and run code. […]<\/p>\n","protected":false},"author":1,"featured_media":39271,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-39270","post","type-post","status-publish","format-standard","hentry","category-uncategorized"],"post_mailing_queue_ids":[],"_links":{"self":[{"href":"https:\/\/www.duck9.com\/blog\/wp-json\/wp\/v2\/posts\/39270","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.duck9.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.duck9.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.duck9.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.duck9.com\/blog\/wp-json\/wp\/v2\/comments?post=39270"}],"version-history":[{"count":0,"href":"https:\/\/www.duck9.com\/blog\/wp-json\/wp\/v2\/posts\/39270\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.duck9.com\/blog\/wp-json\/wp\/v2\/media\/39271"}],"wp:attachment":[{"href":"https:\/\/www.duck9.com\/blog\/wp-json\/wp\/v2\/media?parent=39270"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.duck9.com\/blog\/wp-json\/wp\/v2\/categories?post=39270"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.duck9.com\/blog\/wp-json\/wp\/v2\/tags?post=39270"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}