ClickUp vs Linear

LLM Invented a Detailed, Nonexistent Company Website

A measurement study of six Chinese LLM APIs found a model confidently describing a nonexistent localized corporate website (airtable.cn), including pricing in RMB and a China-specific privacy policy. The author logged 4,023 valid responses across the corpus, discovered the domain was registered to a private individual and returned HTTP 502, and observed that seven of eight tested brand .cn domains were not owned by the brands. To detect these operationally convincing hallucinations the author recommends lightweight predicate checks on extracted URLs, simple WHOIS/DNS/HTTP lookups, repeated phrasing checks (ask the same factual question multiple ways) and strict provenance metadata for every row. The harness and validation scripts are published under CC BY 4.0 on GitHub.

• The author ran a measurement harness against six Chinese LLM APIs and recorded 4,023 valid responses.
• GLM produced a confident, structured description claiming https://www.airtable.cn/ was Airtable's official China website.
• WHOIS and HTTP checks showed airtable.cn was registered to a private individual (since 2016) and the domain returned HTTP 502.

How to Be Funny in Marketing

This playbook-style newsletter explains why humor is a high-value marketing skill that remains difficult for AI to replicate. It outlines the psychological basis for humor (Benign Violation Theory), two golden rules for marketing humor (be specific to your audience; be surprising), and a practical step-by-step approach for testing and scaling funny content. The piece includes case studies and examples — Ahrefs’ Letaido (an AI tool) used to diagnose traffic drops, a viral LinkedIn 'Viral Post Generator' (2M+ users, acquired two weeks after launch), and Wiz’s April Fools’ 'CISOtopia' stunt (110k+ visitors and a 15–20% uplift to Wiz.io). It also lists high-impact marketing surfaces for humor (social, ads, email, 404s, careers pages, swag) and cautions about sensitive topics, punching down, and overuse.

• Ahrefs introduced Letaido, described as an AI tool that runs analyses using Ahrefs’ live data.
• Ahrefs’ CMO tested Letaido on a case where ClickUp’s blog lost 97.6% of its traffic.
• The author’s Viral Post Generator was used by more than 2 million people and was acquired two weeks after launching.

AI 'Psychosis' and Accelerating AI-Driven Layoffs

Box founder Aaron Levie warned of “AI psychosis,” a management blind spot where executives overconfident in AI push to replace human roles without understanding the work. The article cites ClickUp’s announcement to cut 22% of its workforce and replace roles with AI agents, and notes 2026 tech layoffs have nearly matched all of 2025 before midyear. JudyAI Lab frames this as a requirements-analysis failure: teams skip deep interviews with role incumbents, automating visible steps while missing implicit human judgment, exception handling, and edge-case work. The piece urges AI implementers to involve the people doing the work when designing agent/workflow automation to avoid efficiency losses from over-automation.

• Aaron Levie (Box founder) described a phenomenon he calls "AI psychosis," where executives overestimate AI's ability to replace jobs.
• Collaboration platform ClickUp announced cutting 22% of its workforce and stated it will use AI agents to take over those functions.
• The article reports 2026 tech layoffs have nearly matched all of 2025 before midyear, citing accelerating AI-driven workforce reductions.

AI Research Roundup: Terminal Agents, Cloudflare Traffic, Nanochat

This newsletter edition covers recent AI research and tools. Key items include a paper on terminal agent training with self-improving tasks, a method for compressing agent screen memory, a tool for generating editable 3D scenes, and a model that predicts environment responses. Cloudflare's analysis of 206 million web sessions reveals mixed human-agent control, impacting bot detection. A new C file implementation runs a 744B parameter model efficiently, and a code graph tool supports 150+ languages. Additionally, Karpathy's nanochat project trains a GPT-2-class model for $48, and a benchmark shows Apple's SpeechAnalyzer outperforming Whisper Small. The newsletter also highlights videos on model serving and an internal agent at Linear.

• Cloudflare recorded 206 million Precursor evaluations in a day, showing sessions can shift between human and automated control.
• Colibri runs a 744B parameter MoE model from a single C file, using VRAM, RAM, and NVMe as tiers.
• Karpathy's nanochat trains a GPT-2-class model for about $48 on eight H100s.

Coding sessions: Linear Agent can now set up, run, and test your code

Linear Agent can now set up, run, and test your code before returning its work. That means fewer handoffs and changes that are further along when they come back to you.

AI Team Manifest Can Pass Validation But Be Unsafe

The article explains that JSON Schema structural validation can confirm a manifest's shape but cannot guarantee that an AI team configuration is safe or executable. NexFlow uses YAML manifests mapped to JSON Schemas and performs structural checks; the project also runs bounded semantic reference checks across manifests, but full semantic validation and runtime enforcement (permissions, approval gates, credential isolation, auditing) are distinct layers that a production runtime must provide. The author argues tools should report which validation layers they ran and which guarantees remain unverified to avoid misleading users with a single "valid" indicator.

• NexFlow describes AI developer teams through YAML manifests that map each supported manifest kind to a JSON Schema.
• At the reviewed repository checkpoint, `npm run validate` validates 113 manifests against 17 schemas.
• The repository includes a bounded semantic reference smoke check; `npm run semantic-smoke` passes for seven example projects at the same checkpoint.