Reposted by bdougieAll Things Open @allthingsopen.bsky.social · 10/09/2026Brian Douglas (@bizza.pizza), CEO for The Paper Compute Company, joins us at #AllThingsOpen to present "The Beginners Guide to Training Your Own AI Coding Model on Your Own Data"! 2026.allthingsopen.org/sessions/the... 031
bdougie @bizza.pizza · 03/09/2026I swapped Claude for open weights in my agent heist. Claude read the ROM to explain why the move would not work. Qwen walked over and tried the handle. Every seat but one runs on my GPU now. briandouglas.me/posts/2026/0...briandouglas.meEvery Agent Session Should Be a HeistI stopped asking one model to play every role. Pokémon Red showed me how to cast a crew by benchmark and route each job with vLLM. 020
bdougie @bizza.pizza · 09/07/2026My OpenClaw was spending $80 a day to read my email. One config line brought it down to a few dollars. Wrote up where the money went and how observability made the fix obvious: briandouglas.me/posts/2026/0...briandouglas.meMy OpenClaw Was Spending $80 a Day to Read My EmailMy OpenClaw chief-of-staff agent was burning $80 a day in Anthropic API costs. The fix was one config line, and the only reason I found it was observability. 062
bdougie @bizza.pizza · 19/06/2026git push origin origin thenewstack.io/cursor-origi...thenewstack.ioCursor, GitLab and Zed agree GitHub is breaking. They disagree on how to rebuild it.Cursor's Origin, GitLab's Project Switch and Zed's DeltaDB are racing to rebuild code hosting for AI agents as GitHub buckles under the load. 020
bdougie @bizza.pizza · 10/06/2026Claude code shared Fable is storing for sessions for 30 days because it’s useful for understand the safety of the model. My suggestion is you should save your sessions too. If you are not today, put it on paper. Here’s my AI Devcon talk on the subject lnkd.in/efQRKGTxlnkd.inLinkedInThis link will take you to a page that’s not on LinkedIn 010
bdougie @bizza.pizza · 09/06/2026@andrewnez.bsky.social want to join us on a podcast to talk ecosystems? 000
bdougie @bizza.pizza · 20/05/2026I tried training a model instead of writing a skill and have some honest results. briandouglas.me/posts/2026/0...briandouglas.meI Trained a Model Instead of Writing a SkillSkills are a great way to teach an agent a pattern. I got curious whether the agent could just learn the pattern, so I tried training a model while also figuring out what RL actually is. 2163
Reposted by bdougieJohn McBride @johncodes.com · 19/05/2026We out here London 📼📼📼 !!! @bizza.pizza 171
Reposted by bdougieJohn McBride @johncodes.com · 15/05/2026🎤 New episode of Open Source Ready with @bizza.pizza - we talked to @stormypeters.bsky.social about open source at Amazon, how the relationship between AI and open source is actively evolving, and the growing challenges that come with that. Check it out! www.heavybit.com/library/podc...heavybit.comOpen Source Ready | Ep. #37, Is AI Killing Open Source Software? with Stormy Peters | HeavybitOn episode 37 of Open Source Ready, Brian and John speak with Stormy Peters about the evolving relationship between AI and open source software. 011
bdougie @bizza.pizza · 31/03/2026New company launched. papercompute.com/blog/introdu...papercompute.comIntroducing Paper Compute Co: Infrastructure for Running AI Agents in Production - Paper Compute BlogWe built Paper Compute to close the Agent Gap — the inability to prove what an agent did, audit its actions, or replay its decisions. Here's what we built and why. 020
bdougie @bizza.pizza · 15/03/2026They all had the same target in this case. Merge conflicts weren’t an issue since both tests were in no overlapping sections of the code and docs. This is evolution of my previous pokemon experiment briandouglas.me/posts/2026/0...briandouglas.meWhat I Learned Running 10 Pokemon Bots in 36 SecondsHow a DeepMind paper turned my Pokemon agent from 'watch and tweak' into 'run and measure.' 000
bdougie @bizza.pizza · 15/03/2026Raising kids in the bay is fun. Basically like a dog that speaks English eventually 020
bdougie @bizza.pizza · 15/03/2026I just opened source my experience with sweeper agents github.com/papercompute...github.comGitHub - papercomputeco/sweeper: 🧹🧹. Contribute to papercomputeco/sweeper development by creating an account on GitHub. 120
Reposted by bdougieJacob Stordahl but Haunted 🎃 @stordahl.dev · 14/03/2026Hacking on an agent orchestrator that is essentially a bunch of open source tools I like (Git, tmux, opencode and stereOS), duct taped together using Bun. The result provides an isolated workspace in a tmux session you can attach to do work, and a VM where the agent can access the workspace 4192
bdougie @bizza.pizza · 15/03/2026I had 1,992 ESLint errors. Dispatched a sweeper agent to produce 5 parallel Claude Code sessions. Three rounds, 100% fix rate, 54 minutes. briandouglas.me/posts/2026/0...briandouglas.me842 Lint Errors, 5 Parallel Agents, 54 MinutesI built a tool that fans out parallel Claude Code agents to fix lint errors across a codebase. It taught me that the hard problems aren't prompts or models. They're isolation, observability, and memor... 141
bdougie @bizza.pizza · 11/03/2026Blog post covers the approach. Replace pokemon with servers and production data and that’s the focus for customers. I wasn’t aware of cortex cli but going to play around with it once I stand up a snowflake to test in. papercompute.com/blog/agent-p...papercompute.comAgent Played Pokemon for 1,000 Turns. It Never Left the Bedroom - Paper Compute BlogAn AI agent played Pokemon Red for 1,000 turns without leaving the bedroom. The failure wasn't the model. It was the absence of telemetry. 010
bdougie @bizza.pizza · 11/03/2026This is an example of what I am building now with Kafka. github.com/papercompute... I just started working on the sink storage and snowflake is top of mindgithub.comGitHub - papercomputeco/pokemon-kafka: Autonomous Pokemon Red player that reads game memory, makes strategic decisions, and plays headlessly inside a stereOS VM, now with KafkaAutonomous Pokemon Red player that reads game memory, makes strategic decisions, and plays headlessly inside a stereOS VM, now with Kafka - papercomputeco/pokemon-kafka 110
bdougie @bizza.pizza · 10/03/2026This looks cool. Working on something related with Kafka, want to see a demo? 110
bdougie @bizza.pizza · 10/03/2026I used observational memory to train my Pokémon agent. briandouglas.me/posts/2026/0...briandouglas.meWhat I Learned Running 10 Pokemon Bots in 36 SecondsHow a DeepMind paper turned my Pokemon agent from 'watch and tweak' into 'run and measure.' 031
bdougie @bizza.pizza · 04/03/2026Claude Code crashed an hour into a session. No recovery, no history. But I had tapes running. It stores every LLM interaction in a local SQLite db. Pointed Claude at it, recovered everything, and shipped the feature. briandouglas.me/posts/2026/03/04/claude-failed-mid-session-tapes-brought-it-back/ 180
bdougie @bizza.pizza · 02/02/2026So far it has read only access to my GitHub. I’ve been doing some feature planning in a discord chat. Probably or slowing given limited access to edit issues and project management today 030
bdougie @bizza.pizza · 02/02/2026"Setting Up OpenClaw on exe.dev with Discord" #DEVCommunity dev.to/bdougieyo/se...dev.toSetting Up OpenClaw on exe.dev with DiscordI'll be honest—when I first heard about Clawdbot (now rebranded as OpenClaw) a few weeks back, I had... 130
Reposted by bdougieJohn McBride @johncodes.com · 31/01/2026📼 Introducing tapes: transparent AI agentic telemetry. In every single AI session or workflow, there's an incredible amount of lost information on the "what", "why", and "how". www.youtube.com/watch?v=ATeU...youtube.comI started a company.YouTube video by John McBride 281
bdougie @bizza.pizza · 31/01/2026Save open source. Star my repo github.com/papercompute...github.comGitHub - papercomputeco/tapes: Transparent telemetry collector for agentsTransparent telemetry collector for agents. Contribute to papercomputeco/tapes development by creating an account on GitHub. 031
bdougie @bizza.pizza · 31/01/2026What does long term care for agents look like? Wrote about tapes, an open source project @johncodes.com and I have been working on these past 2 weeks. briandouglas.me/posts/2026/0... 071
Reposted by bdougieHeavybit @heavybit.com · 29/01/2026Is product management making a comeback? 🤔 We checked in with development experts Amit Eyal Govrin, @bizza.pizza, and Alex Kurilin to dig into why and what it means to be a good product manager or product engineer in the age of AI. www.heavybit.com/library/arti...heavybit.comWhat Makes a “Good” AI Product Manager? | HeavybitWhat does product management mean in the age of AI codegen, and what does a good AI PM do? Veteran dev leaders weigh in. 011
Reposted by bdougieJohn McBride @johncodes.com · 28/01/2026"There is no secure ai enclave, Neo" I continue to bang the gong on how crazy it is we're shipping and integrating all these agentic systems with critical services with nearly zero guardrails. johncodes.com/archive/2026...johncodes.comthere is no secure ai enclavethere is no secure ai enclave 161
bdougie @bizza.pizza · 18/01/2026I made the joke that every Ralph needs a Super Nintendo and no one laughed, so I wrote post on it. briandouglas.me/posts/2026/0... 3993
bdougie @bizza.pizza · 12/01/2026Had this happen with my .env before sessions was a thing. I asked it to clean up the file and it did. Good to know this trick though. 020
Reposted by bdougieJohn McBride @johncodes.com · 11/01/2026Wrote about the sheer amount of software that is going to be created: the software Cambrian explosion is upon us. Prepare accordingly. johncodes.com/archive/2026...johncodes.comThe software Cambrian explosion.The software Cambrian explosion. 1284
bdougie @bizza.pizza · 18/12/2025Back at 98 in performance thanks to this one trick. briandouglas.me/posts/2025/1...briandouglas.meI Tried Edge SSR in My React Vite AppHow moving to edge functions improved performance without a full rewrite. 010
Reposted by bdougieNick Taylor @nickyt.online · 09/12/2025This was a lot of fun interviewing @bizza.pizza! Thanks for letting me guest co-host @whiskey.fm! whiskey.fm/are-develope...whiskey.fmAre Developers Overthinking Everything? w/ Bdougie - Whiskey Web and Whatnot: Web Development, Neat - Episode 221Recorded live at All Things Open, Robbie and guest co-host Nick Taylor talk with Bdougie (Brian Douglas) about open source’s impact on their careers, the legacy of OpenSauced, and why developer tools ... 162
bdougie @bizza.pizza · 09/11/2025I watched this whole movie waiting to see Sean Penn. Turns out this is not Bill from Bill and Ted. 010
bdougie @bizza.pizza · 27/10/2025I think I made a code rabbit clone. youtu.be/vEoABlyfwq0?...youtu.beI built my own AI code review tool called CodebunnyYouTube video by bdougie 031
Reposted by bdougieCodeTV @codetv.dev · 21/10/2025Now streaming for everyone: build a web-based hotline. Watch @bizza.pizza, Pedro, @jherr.dev, Sherm, Avi, and Dibs code up custom call-in services using Twilio's ConversationRelay. Full episode: codetv.link/wdc/s2e11 131
bdougie @bizza.pizza · 21/10/2025Created a follow up pr for review logging github.com/bdougie/code...github.comfeat: implement review history tracking with artifacts by bdougie · Pull Request #2 · bdougie/codebunnyOverview Implements persistent review history tracking using GitHub Actions artifacts as proposed in #1. What's New Review History Tracking Each review is saved as a GitHub Actions artifact fo... 020
bdougie @bizza.pizza · 21/10/2025github.com/bdougie/code...github.comGitHub - bdougie/codebunny: AI-powered code reviews using Continue AgentAI-powered code reviews using Continue Agent. Contribute to bdougie/codebunny development by creating an account on GitHub. 010
bdougie @bizza.pizza · 21/10/2025I wrote a post on how I use @continue.dev config to own my AI code review data. Looking forward to improving AI with code reviews I can fine tune. briandouglas.me/posts/2025/1...briandouglas.meCodeBunny: Privacy-First AI Code Reviews with ContinueAI code reviews that run in your GitHub Actions environment, keeping your code in your repository. Built on Continue for flexible, privacy-focused code analysis. 160
