Creeta — AI developer tools & ecosystem news
17k tokens → 1.4k — Headroom keeps the originals retrievable
Open-source context compression middleware for agent pipelines: 60–95% token cuts, CCR reversibility, AST-aware engines.
Cognition's $26B needs $1B ARR by December. The math is tight.
$26B valuation on $492M ARR: Cognition's Series D metrics, the Windsurf attribution question, and the $1B ARR target.
Booed at graduation — the AI skeptics you'll be shipping to
MIT Technology Review's May 2026 Hype Index covers graduation boos, Gen Z sentiment (46%), and record AI fundraising.
Opus 4.8 kills budget_tokens — here's what else moved
Opus 4.8: fast mode, mid-session system prompts, 1K cache floor. Old budget_tokens syntax returns 400.
llama-bench skipped FA on capable GPUs — b9437 corrects it
llama.cpp b9437 (May 30): -fa goes auto, -ngl to -1 in llama-bench. Your pre-b9437 comparisons need a flag audit.
Qwen3.6-35B NVFP4 runs on one H100 — A100 owners are out
FP4-quantized Qwen3.6-35B fits in ~23 GB on Hopper. vLLM serve commands, env vars, DGX Spark config, and gotchas.
Step 3.7 Flash is a drop-in — except for one endpoint detail
StepFun Step 3.7 Flash: 198B MoE with native vision, Advisor Mode, and an OpenAI-compatible API you can call today. Includes endpoint gotchas and reasoning_effort examples.
You don't pick the RL algorithm — SIA's Feedback loop does
SIA co-evolves scaffold and LoRA weights in one loop. Install, run LawBench, and add custom evals — Hexo Labs, May 2026.
NVIDIA cut Qwen3.6-35B 3×. Accuracy barely moved.
NVIDIA's NVFP4 Qwen3.6-35B checkpoint on HuggingFace: 3.06× memory reduction, <1% accuracy loss, Blackwell-native, vLLM flags included.
Overslash holds the credentials. Your AI only gets a handle.
Overslash injects secrets by handle at the gateway, limits blast radius per agent, and escalates out-of-scope calls to human approval. Free self-hosted or €3/seat cloud.
Harness edits slowed the GPU kernel 10×. Weights fixed it.
SIA edits its scaffold and fine-tunes weights via LoRA — 70.1% LawBench, 12.4% faster GPU kernels, MIT-licensed.
Omni skips the re-render — nine demos show the difference
Gemini Omni and 3.5 Flash demo breakdown: nine I/O 2026 clips, scene-preservation vs parallel coding, API availability.
AI developer tools and ecosystem news for developers and technical founders