Shane Caldwell https://hackbot.dad/ Recent content on Shane Caldwell Shane Caldwell https://hackbot.dad/ https://hackbot.dad/ Hugo -- 0.146.2 en-us Sat, 14 Mar 2026 00:00:00 +0000 The Tests All Pass https://hackbot.dad/writing/tests-all-pass/ Sat, 14 Mar 2026 00:00:00 +0000 https://hackbot.dad/writing/tests-all-pass/ METR's SWE-bench analysis shows us taste isn't verifiable. Intro to GPUs For the Researcher https://hackbot.dad/writing/intro-to-gpus/ Mon, 05 Jan 2026 00:00:00 +0000 https://hackbot.dad/writing/intro-to-gpus/ Getting comfortable with the hardware on a quest for more MFU. All Reduce Across the Atlantic: Bandwidth in Decentralized Training https://hackbot.dad/writing/all-reduce-across-atlantic/ Fri, 12 Dec 2025 00:00:00 +0000 https://hackbot.dad/writing/all-reduce-across-atlantic/ The practical realities of devestatingly high communication cost in training. Twenty Billion Tokens of What, Exactly? https://hackbot.dad/writing/20b-tokens-of-what/ Mon, 01 Dec 2025 00:00:00 +0000 https://hackbot.dad/writing/20b-tokens-of-what/ Looking at the data and letting it look back at us. Pretraining at home: 20B tokens from 222 hours to 12 https://hackbot.dad/writing/pretraining-at-home/ Sun, 23 Nov 2025 00:00:00 +0000 https://hackbot.dad/writing/pretraining-at-home/ Optimizing training a Llama 3.2 1B model so we can pretrain in a day without going broke. Offsec Evals: Growing Up In The Dark Forest https://hackbot.dad/writing/offsec-evals-dark-forest/ Tue, 28 Oct 2025 00:00:00 +0000 https://hackbot.dad/writing/offsec-evals-dark-forest/ If you contribute a public benchmark, are you giving free capability to your competitors? DiLoCo: Data Parallelism for the Datacenter Poor https://hackbot.dad/writing/data-parallelism-for-the-poor/ Fri, 03 Oct 2025 00:00:00 +0000 https://hackbot.dad/writing/data-parallelism-for-the-poor/ Distributed training sans datacenter. RL Needed LLMs Because Agency Requires Priors https://hackbot.dad/writing/rl-llms-and-priors/ Tue, 19 Aug 2025 00:00:00 +0000 https://hackbot.dad/writing/rl-llms-and-priors/ We tried RL once. It didn't work. I'm confident it will this time. GPT-5 is Good, Actually: The Agony and Ecstasy of Public Benchmarks https://hackbot.dad/writing/agony-and-ecstasy-evals/ Sun, 17 Aug 2025 00:00:00 +0000 https://hackbot.dad/writing/agony-and-ecstasy-evals/ An attempt to explain why benchmarks are either bad or secret, and why the bar charts don't matter so much. The Religious Devotion of Haskell https://hackbot.dad/writing/haskell-empathy/ Mon, 01 Jul 2024 00:00:00 +0000 https://hackbot.dad/writing/haskell-empathy/ An exploration of functional programming through Haskell, motivated by trying to understand the near-religious devotion its practitioners have for the language. The Input Sanitization Perspective on Prompt Injection https://hackbot.dad/writing/prompt-injection/ Sun, 02 Jul 2023 00:00:00 +0000 https://hackbot.dad/writing/prompt-injection/ An analysis of prompt injection vulnerabilities in large language models and why they represent a fundamental security challenge. Infosec's Data Problem https://hackbot.dad/writing/infosecs-data-problem/ Thu, 02 Jun 2022 00:00:00 +0000 https://hackbot.dad/writing/infosecs-data-problem/ Exploring the fundamental data sharing challenges that limit machine learning progress in information security, and why the field needs its own ImageNet moment. Deep Reinforcement Learning for Security: Toward an Autonomous Pentesting Agent https://hackbot.dad/writing/towards-autonomous-pentesting/ Tue, 28 Apr 2020 00:00:00 +0000 https://hackbot.dad/writing/towards-autonomous-pentesting/ An exploration of using deep reinforcement learning to create autonomous penetration testing agents, examining the challenges and potential solutions for automating cybersecurity assessments. An ML Eng's Review of OSCP https://hackbot.dad/writing/oscp-review/ Sun, 26 Apr 2020 00:00:00 +0000 https://hackbot.dad/writing/oscp-review/ A comprehensive review of the Offensive Security Certified Professional (OSCP) certification from the perspective of a machine learning engineer entering the security field.