tessera Tech, tiled.
TUE 14.07.2026 · 00:20 UTC EDITION T-2026-W29

Category / T-CAT / SOFTWARE

Software

Articles58
This week10
Last updatedJuly 13, 2026

Software / T-2026-5069

llama.cpp b9775: The quiet infrastructure of local AI

llama.cpp b9775 ships a fix for speculative decoding errors. The real news is what the release tells us about the state of local AI infrastructure.

Sourceggerganov/llama.cpp b9775

Software / T-2026-5734

The 83 tips that reveal how AI agents actually work

A community repo of 83 tips for Claude Code reveals the gap between vibe coding and agentic engineering. The details matter more than the hype.

Sourceshanraisshan/claude-code-best-practice

Software / T-2026-4458

HeyGen's HyperFrames bets plain HTML beats React for AI video

HeyGen open-sources HyperFrames, an HTML-native video renderer built for AI agents, challenging Remotion's React-based approach.

Sourceheygen-com/hyperframes

Software / T-2026-9757

The system prompt leak repo that changed how we see AI models

A GitHub repository collects leaked system prompts from Claude, ChatGPT, Gemini, and dozens of other AI chatbots, exposing the hidden rules that govern every response.

Sourceasgeirtj/system_prompts_leaks

Software / T-2026-7350

jcode: A coding agent harness that treats resource efficiency as a feature

The open-source jcode harness uses 27.8 MB per session and boots in 14 ms. Claude Code uses 386.6 MB and takes 3.4 seconds. The gap is a design choice.

Source1jehuang/jcode

Software / T-2026-3264

BuilderIO's Agent-Native framework fuses UI and agent into one app

BuilderIO's Agent-Native framework treats the agent and the user interface as equal citizens, sharing one database and one action surface.

SourceBuilderIO/agent-native

Software / T-2026-9913

Hyper-Extract turns any document into a knowledge graph with one CLI command

Hyper-Extract wraps LLM extraction into 80+ YAML templates, supporting graphs, hypergraphs, and spatio-temporal structures — all from one CLI command.

Sourceyifanfeng97/Hyper-Extract

Software / T-2026-3271

OpenMontage turns AI coding assistants into video production studios

The open-source OpenMontage system lets AI coding assistants produce complete videos from a single prompt, with costs as low as $0.15 per clip.

Sourcecalesthio/OpenMontage

Software / T-2026-4571

Trycua’s Cua stack gives AI agents a full desktop OS to control

Trycua releases Cua, an open-source stack for building, benchmarking, and deploying AI agents that control full desktop operating systems in the background.

Sourcetrycua/cua

Software / T-2026-3986

llama.cpp b9637 adds native Cohere2MoE support, shrinking the gap between open-source and proprietary inference

llama.cpp b9637 ships a dedicated Cohere2MoE parser, bringing native support for North Code models to the open-source inference stack. The move highlights how MoE architecture is…

Sourceggerganov/llama.cpp b9637

Software / T-2026-1845

Andrew Ng's aisuite is a bet that AI needs a universal adapter layer

Andrew Ng's aisuite wraps dozens of LLM providers behind a single API and adds an agents framework. The bet is that developer friction, not model capability, is the bottleneck.

Sourceandrewyng/aisuite

Software / T-2026-3956

llama.cpp b9628 ships SYCL support, broadening GPU access beyond CUDA

llama.cpp b9628 adds SYCL support for Intel GPUs, a quiet but significant step toward breaking NVIDIA's lock on AI inference hardware.

Sourceggerganov/llama.cpp b9628