TLDR AI 2026-06-22

Unwrap is the AI-powered customer intelligence platform trusted at scale by Stripe, Southwest Airlines, Perplexity, WHOOP and other customer-centric companies.

With Unwrap, you get:

• All customer feedback automatically categorized by AI + NLP
• Structured feedback you can query using Unwrap Assistant, or in your favorite tools using Unwrap's MCP
• Real-time alerts for new feedback that demands your attention
• A clear view of customer sentiment

👉 Unwrap is offering a free trial of its tools to TLDR subscribers. Grab time with the team to get set up

Sakana Fugu is a multi-agent system that behaves like a single model. Fugu can decide whether to handle requests directly or coordinate a team of expert models. It manages model selection, delegation, verification, and synthesis. Users simply call one model, and a coordinated system of experts does the work. Sakana Fugu and Fugu Ultra are available today through a single OpenAI-compatible API.

Mercury 2 is a reasoning language model that generates about 1,000 tokens per second. It uses diffusion, the same trick that turns static into a photo in image generators like Stable Diffusion. The model is best for speed-sensitive, high-volume parts of workflows rather than the hardest frontier reasoning. It is only available via API/cloud.

Nobel laureate John Jumper is leaving DeepMind for Anthropic after nine years. Jumper, who co-led the AlphaFold team, won a Nobel Prize for predicting protein structures. His departure follows struggles at DeepMind in selling coding tools to businesses.

A transparency audit found that DiffusionGemma remained similarly monitorable to Gemma despite its diffusion-based architecture. The analysis highlighted a gap between variable transparency and algorithmic transparency, and explored phenomena such as non-chronological reasoning, token smearing, and intermediate-context reasoning.

Claude Fable 5 and Claude Mythos 5 were shut down by the White House via an imposition of export controls. The Trump Administration said it was due to a jailbreak of Fable - this turned out to just be saying 'fix this code'. Anthropic has been told to fix this 'jailbreak', which is impossible. It's now been over a week since the pause in deployment and the situation has yet to improve.

Token costs scale with agents, but are they tied to value? This Glean whitepaper explains how high-quality context and indexed retrieval reduce unnecessary reasoning, lower token consumption, and improve results, and how intelligent model routing helps enterprise AI systems do more useful work per token. Download the whitepaper →

NVIDIA ENPIRE is a closed-loop framework that enables coding agents to iteratively improve real-world robot policies through automated resets, evaluation, verification, and refinement.

AI coding workflows are shifting from prompt engineering to loop engineering, where developers build systems that repeatedly prompt, evaluate, and re-prompt agents until a measurable goal is achieved.

Morph LLM optimizes open coding models by training a drafter on coding output, not the internet, for faster speculative decoding, achieving a 3.07x speedup. Autoresearch automates kernel tuning for low-demand GPUs like NVIDIA and AMD, enhancing warp-decode kernels to reach 162 tok/s on affordable hardware. Interconnect over PCIe replaces expensive NVLink using custom kernels, maintaining performance by sharing caches via TCP, cutting time-to-first-token by 84%.

We're hiring a senior PMM to own product marketing at TLDR. You'll define our positioning, build out sales enablement, and lead every launch. Learn more.

A large part of the job search journey involves managing all of the emotions that come with being on the market. There's a lot of social perception to navigate. It can be stressful navigating a huge decision space with incomplete information where small choices have an outsized impact. This post takes a look at what the job search experience is like right now.

A speculative thought experiment called Europe 2031, penned by Brussels-based thinktankers, paints a picture of a world where Europe's lack of investment into datacenters results in it being far behind the US and China. The lack of investment results in Europe's economy being in shambles as it lacks its own AI. In this world, populism is surging, the Euro is wobbling, and cyber-attacks are shredding EU businesses. The scenario was read by members of the European parliament and brought up in discussions between British and German officials last week.

Attention might be all you need, but modern models use many different variants of it.

Tech giants, including Google, Microsoft, and Cisco, are developing the Agentic Resource Discovery (ARD) protocol.

Companies led by highly ideological founders eventually trigger resistance from regulators, investors, customers, and governments as their influence grows beyond traditional institutional controls.

New safeguards on recent models make the idea of running open models on private machines much more appealing.

Linear A is a Bronze Age writing system that appeared sometime around 1800 BC and was used until 1450 BC.