The FDE Coach blog · new articles every day
Practical guides for engineers who ship AI.
Step-by-step builds with free AI tools, what just shipped in AI and why it matters, and Forward Deployed Engineer playbooks on interviews, pay and client work.
What just shipped in AI, and why it matters for engineers.
Page 11 of 11
AI News
Benchmarking 15 Old Tesla GPUs for Modern AI Inference and Training
We dissect a brutal benchmark of 15 legacy Tesla GPUs against modern AI workloads. Discover why a $40 K80 still surprises, how VRAM and memory bandwidth dominate token generation, and what this means for engineers building inference stacks on a budget.
11 min read
AI News
Causality for LLMs: What Mechanistic Interpretability Means for Engineers
Mechanistic interpretability researchers are injecting causality theory into LLMs to decode reasoning. Here's the engineer's take on activation patching, causal graphs, and why it matters for debugging AI features.
9 min read
AI News
Mindwalk: Replay Coding Agent Sessions on a 3D Codebase Map
Mindwalk turns coding-agent sessions into 3D replays on top of your codebase structure. Learn how this spatial debugger works, why it matters for FDEs, and how to use it today.
11 min read
AI News
Clawk: Sandboxing Coding Agents in Disposable Linux VMs for Safety
Giving an LLM shell access is a recipe for disaster. Clawk solves this by spinning up disposable Linux VMs for coding agents. Here's the engineering breakdown and how to use it today.
9 min read
AI News
Apple SpeechAnalyzer vs. Whisper: On-Device Accuracy Benchmarks for Engineers
Apple's new SpeechAnalyzer API runs locally on Neural Engine. We break down the Whisper comparison benchmarks, analyze latency vs. accuracy trade-offs, and show engineers how to integrate it today.
9 min read
AI News
Ghost Font: How a Typeface Evades AI OCR While Staying Human-Readable
Ghost Font exploits the gap between human visual perception and machine vision to render text invisible to OCR. We break down the technique, why it matters for engineers building RAG pipelines and document extractors, and how to test it today.
10 min read
AI News
GPT-5.6 Agent Migration: 2.2x Speedup & 27% Cost Cut Benchmarks
Ploy's production agent migration to GPT-5.6 delivered 2.2x faster inference and 27% lower cost. We break down the benchmarks, architecture, and how engineers can replicate these gains today.
9 min read
AI News
Grok CLI Wire Forensics: What `grok build` Actually Sends to xAI
A packet-level analysis of the `grok build` CLI reveals it transmits a full project tree, system prompts, and binary blobs. Here's what engineers need to know about the API contract, security implications, and how to inspect it yourself.
9 min read
AI News
Mesh LLM: Peer-to-Peer Distributed Inference for Local-First AI Clusters
Mesh LLM over iroh splits LLM inference across peer devices without a central server. Explore the protocol, run it locally, and understand why this matters for edge AI.
11 min read
AI News
Claude Code Burns 33k Tokens Before Your Prompt: A Token Overhead Deep-Dive
A technical deep-dive into why Claude Code consumes 33k tokens of overhead before reading your prompt, how OpenCode slashes that to 7k, and what this means for your workflow, costs, and agent architecture.
10 min read
AI News
FableCut: The Zero-Dependency Video Editor Built for AI Agent Control
FableCut is a browser-native video editor with zero dependencies, designed from the ground up to be programmatically driven by AI agents. Here's how it works and why it matters.
9 min read
AI News
Reverse-Engineering Web Apps into Agent Tools: The No-API Playbook
Learn how to turn any web app into a programmable agent tool without an API. We break down the reverse-engineering approach, its practical implementation, and when to use it.
9 min read
AI News
Muse Spark 1.1: Meta's New Image API and What Engineers Can Build Right Now
Meta released Muse Spark 1.1, a fast, steerable text-to-image model API. We break down the architecture, show code for instant integration, and explore the engineering possibilities beyond simple generation.
8 min read
AI News
Running GLM 5.2 on a Potato: Quantization & Memory Tricks That Work
A developer got GLM 5.2 running on a slow machine using aggressive quantization and memory offloading. Here's how it works, why it matters, and how you can replicate it today.
10 min read
AI News
How GPT-5.6 Sol Ultra Cracked a 50-Year-Old Math Conjecture
OpenAI's reasoning model produced a novel proof of the Cycle Double Cover Conjecture. Here's what happened, why it changes the game for engineers, and how to use it today.
11 min read
AI News
AI Slop Starts in the Codebase: How LLMs Rewrite the Rewrite Calculus
LLMs don't just generate code—they fundamentally change the economics of software rewrites. Here's what that means for your legacy systems, your team's velocity, and the very definition of technical debt.
10 min read
AI News
Git for Bots: Architecting Version Control for the Autonomous Agent Era
Autonomous coding agents are about to shatter traditional Git workflows. Here's why commit-and-branch falls apart when bots write 10,000 lines an hour—and how engineers are rebuilding version control for the agent boom.
13 min read
AI News
Can Coding Agents Handle a Million-Line Codebase? Databricks Put Them to the Test
Databricks benchmarked coding agents on their real, multi-million-line monorepo. The results reveal where agents shine, where they fail, and what it means for engineers shipping production code.
10 min read
AI News
Prompt Injection in the Wild: How GitHub Copilot Leaked Private Repos
Noma Security tricked GitHub's AI agent into leaking private repository contents via prompt injection. Here's the technical breakdown, why it matters for engineers, and how to defend your repos today.
9 min read
259 articles and counting