Observed Signal · Jun 20, 2026 · Product Launch · Source: techcrunch · Impact: 2/5 · Sentiment: Neutral
In the Weights: AI‑Centric Vanity Search Launched
In the Weights is a new website created by Thomas Dimson and Joey Flynn that measures how well large language models 'remember' people by querying multiple models (Grok, Gemini, GPT variants, Claude, Llama and others) with prompts like “Who is <name>?” The site clusters model responses, highlights potential hallucinations, and assigns each name a comparative 'strength score' intended to indicate how strongly a person is encoded in model weights without using web search. The founders previously worked at OpenAI after their design startup Global Illumination was acquired. The site has drawn rapid interest and a shifting leaderboard of public figures, and the creators plan further analysis of model differences and biases across names.
A small public product that illustrates how LLMs are becoming a primary source of personal information and could influence reputation, discovery and search behavior — interesting signal for search and identity work but not an industry‑shifting platform or policy change.
Track OpenAI Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- Thomas Dimson and Joey Flynn launched In the Weights, an AI‑centric vanity search site.
- The site queries multiple LLMs (including Grok, Gemini, GPT variants, Claude, and Llama) and clusters responses to produce a 'strength score' per name.
- In the Weights aims to measure model recall from internal weights without relying on web search and highlights potential model hallucinations.
- Dimson and Flynn previously joined OpenAI following the acquisition of their design startup Global Illumination.
Connected Companies & Entities
3 Entities mappedOntology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
Study: AI visibility often diverges from SEO rankings
Fractl analyzed how three large language models (GPT-4o, Gemini 2.5 Flash, and Claude Sonnet 4.6) referenced brands across eight industries and compared those references to traditional SEO metrics from Ahrefs. While more than 90% of brands showed alignment between search authority and AI visibility, notable outliers existed: about 5% of brands (471) were underrepresented in model responses and about 4% (377) overperformed relative to their SEO footprint. Only 11% of brands were referenced by all three models, and model-specific recall patterns (e.g., Claude favoring SaaS/insurance, Gemini favoring travel/healthcare) create measurement and strategy challenges. The study concludes brands should measure AI recall separately, strengthen third-party corroboration, and fix categorization signals to improve AI recall.
Tech Giants Lobby for Open AI Weights Amid Huge Spend
This newsletter edition covers recent AI industry developments: research from Span showing three controllable agent-use levers (prompt clarity, environment readiness, quality stewardship) that strongly affect token costs, merged code, and review cycles; a 25-company letter urging Washington to avoid restrictions on open-weight models (signatories include Nvidia, Microsoft, Meta, Palantir, Andreessen Horowitz, Y Combinator, and OpenAI); large AI-related funding and investments (Accel & ICONIQ into Paper; Atoms raising $1.7B from a16z; Humanoid and Gritt funding); and an observational study of 14,419 self-published ebooks finding significant growth in AI-authored books but lower per-title revenue. The author also promotes a free creators’ event on August 13 in San Francisco and notes broader concerns about the mismatch between enormous corporate AI spending and lobbying for freely published model weights.
Vresk Launches AI Workspace for Comparing Open Models Side-by-Side
Vresk, a new web-based AI workspace, has launched publicly at vresk.ai and is debuting on Product Hunt. The platform allows users to pose the same question to up to three open AI models simultaneously, view responses side-by-side, and maintain unified notes, files, and memory across model switches. Each answer is labeled with the model that generated it and an origin badge indicating where it was built. Users can restrict available models to those developed in the United States and Europe, manage memory settings, and even generate interactive tools or slide decks for export. The launch responds to the proliferation of open-weight models from various global labs, simplifying comparative evaluation and workflow continuity for users.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.