Daily Feed — 2026-04-01

This content is AI-generated by my RSS reader tool. Summaries and novelty ratings should be taken with a pinch of salt.

Local, CPU-Friendly, High-Quality TTS (Text-to-Speech) with Kokoro

Source: ariya.io | Tags: cpu-friendly, kokoro, local, speech-synthesis, tts | Published: 2026-04-01 | Novelty: 38%

The article highlights Kokoro, an 82M parameter TTS model that generates high-quality speech on CPUs, making it suitable for local use without compromising privacy. It provides a simple Docker container setup and example code in JavaScript and Python to integrate the TTS functionality into applications. The synthesis speed ranges from 1.5 seconds (AMD Ryzen) to 4.7 seconds (Intel Core), demonstrating its capability even on older CPUs.


carlos@github

Source: Carlos Becker | Tags: career, developers, github | Published: 2026-03-31 | Novelty: 36%

Carlos is joining GitHub after spending his career building tools for developers, including GoReleaser and work at Charm. He will start in a couple of weeks while still on parental leave with his newborn daughter, emphasizing the importance of this opportunity despite challenging circumstances.


Falcon Perception

Source: Hugging Face - Blog | Tags: document-understanding, early-fusion, ocr, perception, transformers | Published: 2026-04-01 | Novelty: 33%

Falcon Perception uses a minimalistic architecture with one backbone for both vision and text processing, employing early-fusion techniques to handle dense object detection and OCR tasks. Key innovations include an autoregressive interface that scales well to hundreds of instances and a separate 0.3B-parameter variant called Falcon OCR for document understanding, outperforming larger models on specific benchmarks while offering higher serving throughput.


Notes from March 2026

Source: Evan Hahn (dot com) | Tags: ai, climate-change, gaming, software | Published: 2026-03-31 | Novelty: 32%

The author published a disclaimer about the use of generative AI in their blog, created a satirical software project called llm-eliza, and shared thoughts on the impact of AI on human creativity. They also fixed a bug in fzf, discovered a standard for asserting human authorship (human.json), and wrote articles on gaming differences and climate change conspiracies.