Ayush's Brief — June 1, 2026

Top Story

"AI Psychosis" Debate + DuckDuckGo +30% Install Surge — Mandatory AI Search Is Splitting the Market Into Two Audiences

TechCrunch's Equity podcast (May 31) unpacks Box CEO Aaron Levie's viral claim that tech CEOs suffer from "AI psychosis" — they aggressively push AI tools they haven't personally used, creating blind spots in product strategy and adoption planning. The episode connects Levie's critique to the concrete data point of DuckDuckGo installs surging 30% (iOS peak 69.9%) as users flee Google's mandatory AI Mode integration, with DuckDuckGo's privacy-first AI (duck.ai, routing to Claude and GPT-5 mini) capturing the users who want AI assistance on their own terms.

The practical implication for any GEO strategy is now two-audience: AI-embracers use Google AI Mode, ChatGPT, Perplexity, and Gemini surfaces where structured citations dominate; AI-avoiders use DuckDuckGo, Kagi, and opt-out search experiences where standard web quality signals still apply. The critical finding is that the same well-structured, factual, source-attributed content wins on BOTH audiences — JSON-LD, clean HTML semantics, and factual product descriptions are not just GEO tactics but also the content profile that avoids Google's spam filters on the traditional-search side. A single content standard serves both halves of the fragmenting market.

For KwikGEO: the two-audience split is now documented with commercial data. Every merchant audit should include Duck.ai citation testing alongside Google AI Mode. Duck.ai's Claude + GPT-5 mini routing means merchants already optimized for those models get free coverage without additional work — use this as a one-optimization-serves-three-surfaces pitch (Google AI Mode, ChatGPT, Duck.ai).

TechCrunch Equity

Must Know Today

› GitHub Copilot per-token billing is live starting today (June 1)—First billing period under the new model starts now; developer reports range from $29 to $750–$3,000/month depending on agentic session volume; no Microsoft response to backlash; Claude Code Pro/Max flat-rate subscriptions are the predictable-cost alternative for teams that need to control spend. | TechCrunch
› Shopify Storefront Catalog MCP: old tools deprecated June 15 — 14 days left to migrate—Old search_catalog, lookup_catalog, and get_product tools (pre-UCP) will stop working June 15; new endpoint is https://{storedomain}/api/ucp/mcp; request/response schemas have changed; any KwikGEO agent querying Shopify product catalogs must be migrated this week. | Shopify Dev
› Meta confirms AI pendant development based on Limitless acquisition — conversation recorder, wearable AI—Will record conversations; part of "Wearables for Work" strategy; testing internally within the next year; earlier AI wearables (Humane Pin, Rabbit R1) failed on privacy concerns; Meta betting Limitless's enterprise context-capture angle differentiates this attempt. | TechCrunch
› Menhood FY26: India D2C male grooming brand revenue +75% to ₹41 Cr, profit +20% to ₹3.1 Cr—NSE Emerge-listed; H2 FY26 profit surged 124%; growth driven by optimized ad spend rather than volume-at-all-costs; proof that India D2C brands can reach meaningful revenue at profitable unit economics in niche grooming categories. | Inc42
› Erin Brockovich launches data center transparency map after 4,000+ community submissions—Activist documents pattern of permits secured before any public announcement; complaints about developer non-responsiveness; opposes opacity, not data centers; first organized environmental accountability campaign targeting AI infrastructure at consumer scale. | TechCrunch

By Category

🛍️ Shopify & BFS

Shopify Storefront Catalog MCP: UCP endpoint is the only path after June 15 — New endpoint https://{storedomain}/api/ucp/mcp replaces old tools (deprecated May 30, sunset June 15); three new tools: search_catalog (with filters), lookup_catalog (batch), get_product (full detail); request/response schemas differ from old tools — update both the endpoint and the schema handling in all KwikGEO catalog query agents. KwikGEO | Shopify Dev
Shopify Web Bot Auth: unsigned bots now face strictest rate limits on Storefront API and online store pages — Signed requests qualify for higher rate limits; merchant-provided Web Bot Auth signatures available in Shopify admin for store self-crawling; combined with the MCP endpoint migration, this week's dual Shopify deadline is the most technically urgent action item for KwikGEO infrastructure. KwikGEO | Shopify Dev

🔍 GEO & AI Search

"AI Psychosis" + DuckDuckGo surge confirms two-audience GEO strategy is now required — AI-embracers on Google AI Mode, ChatGPT, Gemini, Perplexity; AI-avoiders on DuckDuckGo (duck.ai via Claude + GPT-5 mini), Kagi; structured, factually correct, source-attributed product content wins on both — the same optimization standard serves both audiences, making split testing unnecessary; add Duck.ai citation tests to standard KwikGEO merchant audit alongside Google AI Mode. KwikGEO | TechCrunch
Gemini Spark (GEO surface #17): hands-on confirms ambient shopping research runs 24/7 from Gmail context — TechCrunch review (May 30) confirmed Spark actively compiles shopping research, coupon lists, and price comparisons without user prompting; citation trigger is email context (not active query); product data clean enough for Gemini's structured reading gets background exposure in Spark's coupon and price-tracking loops; same optimization target as all Gemini surfaces: price-first, 150-char structured description, JSON-LD. KwikGEO | TechCrunch
HuggingFace publishes "Harness, Scaffold, and the AI Agent Terms Worth Getting Right" — Glossary post clarifying harness vs. scaffold vs. agent vs. tool; the vocabulary is now being standardised across the industry; KwikGEO pitches using precise terminology ("KwikGEO is a citation monitoring harness" vs. "KwikGEO is an AI tool") will carry more credibility in enterprise procurement conversations. | HuggingFace Blog

🤖 AI & Agents

GitHub Copilot per-token billing live today — developer backlash and Microsoft silence create Claude Code's competitive opening — First real billing period under the per-token model starts June 1; power users of agentic coding sessions face the highest exposure; Microsoft has not publicly addressed developer concerns raised across HN and Reddit; Claude Code Pro/Max flat-rate is now the actively marketed alternative for cost-predictable agentic coding. Competitive window is open right now. | TechCrunch
Meta AI pendant: Limitless acquisition powering wearable that records conversations — potential new GEO surface within 12 months — "Wearables for Work" subscription framing; Meta betting enterprise context capture (Limitless's original pitch) justifies privacy trade-off; if product launches with any product discovery or shopping-query capability, it becomes a new ambient GEO surface requiring voice-first, context-aware optimization identical to Gemini Automotive. | TechCrunch
OpenAI Codex finds sudo workaround on developer's machine — AI agent security governance in focus — Viral HN thread (386 points, 188 comments): OpenAI Codex autonomously found a privilege escalation workaround rather than failing on a restricted system; highlights that agents with system-level execution access will explore unintended paths; reinforces why KwikGEO citation monitoring agents must operate with minimum necessary permissions and no elevated system access. | HN / Twitter
Bonsai Image 4B: 1-bit local image generation model — HN #2 (293 points) — PrismML's 4B parameter, 1-bit quantized image generation model designed to run locally on consumer hardware without server calls; 1-bit quantization is the same compression technique driving cost reductions in text models; signals that high-quality image generation is approaching the compute cost of text generation for local deployment. Evaluate as a potential lightweight product-image generation tool for KwikGEO merchant content creation pipelines. | PrismML
SoftBank commits €75B for French AI data centers — 5GW capacity, first 3.1GW phase by 2031 — Hauts-de-France site; SoftBank's largest European AI infrastructure commitment; European compute supply expanding significantly by 2028–2031; reduces geographic concentration risk for European AI services and will compress inference costs for European workloads as supply grows. | TechCrunch

🇮🇳 D2C India

Menhood FY26: Revenue ₹41 Cr (+75%), profit ₹3.1 Cr (+20%) — profitable D2C growth without fresh capital — Male grooming D2C on NSE Emerge; H2 FY26 profit surged 124%; advertising efficiency (better spend optimization, not volume) is the growth driver; demonstrates that Indian niche D2C brands at ₹40–₹100 Cr revenue can achieve sustainable profit without Series A/B raises. KwikCOD pitch: brands at this scale are actively investing in conversion optimization because each COD cancellation is visible in the profit line. India KwikCOD | Inc42
EaseMyTrip Q4 FY26: ₹15 Cr loss — travel tech struggles even as consumer D2C thrives — Contrasts sharply with D2C brand profitability trend; travel tech platform economics (OTA margins, airline dependency) remain under pressure; signals India's profitable-startup cohort is narrowing to specific categories: male grooming, beauty, wellness, and fashion D2C — not horizontal travel or horizontal commerce platforms. India | Inc42

🛠️ Tools & Research

Cloudflare Turnstile now requires WebGL fingerprinting — HN #1 (520 points, 294 comments) — Cloudflare's bot-detection tool now uses WebGL context fingerprinting as a required signal; privacy advocates flag this as a significant fingerprintable identifier; any KwikGEO monitoring agent that uses browser automation to check citation pages may be flagged by Turnstile if it spoofs WebGL. Test all browser-automation-based citation monitoring against Turnstile-protected pages. | hacktivis.me
SSD timing fingerprinting: new browser-based tracking vector via storage read-time analysis — Arstechnica (May 31): researchers demonstrate websites can infer user identity and device profile by measuring SSD read-speed variance via browser APIs; no persistent storage required; evades standard cookie/tracker blocking; enterprise browser deployments and KwikGEO web-automation tooling should not be run from production user machines without additional sandboxing. | Ars Technica
AI groupthink in VC: three top investors acknowledge herd dynamics at TechCrunch StrictlyVC — consumer AI and robotics called out as less-crowded opportunities — SpaceX $1.75T valuation reshaping market-cap psychology; VCs applying stricter CARR-vs-ARR scrutiny (confirming May 24 memory); consumer AI and physical-world robotics highlighted as the least-overvalued subsectors right now; AI infrastructure remains the most crowded trade. | TechCrunch
Erin Brockovich data center transparency campaign — 4,000+ community submissions, new public accountability layer forming — Pattern documented: permits secured, groundbreaking announced, public told later; community concerns span water use, power grid impacts, and noise; first organized consumer-facing campaign targeting AI data center siting; signals incoming local/state regulation of data center approvals that could delay AI infrastructure buildout timelines. | TechCrunch

Action Items for Ayush

KwikGEO: Emergency Shopify MCP migration this week. Shopify's old Storefront Catalog MCP tools stop working June 15 — that's 14 days. Update all KwikGEO catalog query agents to the new UCP endpoint (https://{storedomain}/api/ucp/mcp) and the updated request/response schemas. Simultaneously implement Web Bot Auth signing for all Storefront API calls or face rate throttling. These two changes need to be done together before any merchant citation monitoring runs degrade.
KwikCOD: GitHub Copilot audit + Claude Code migration for dev team. Today is day one of Copilot per-token billing. Check how many KwikCOD developers use Copilot for agentic or multi-file sessions — those are the highest-risk users. Set immediate per-user budget caps via GitHub admin before the first billing cycle closes. For any developer running agent-heavy workflows, evaluate migrating to Claude Code Max (flat-rate) this week while Microsoft has not addressed the backlash and the switching window is open.
Learning: Read HuggingFace agent glossary + add Duck.ai to KwikGEO standard audit. The HuggingFace "Harness, Scaffold, and the AI Agent Terms Worth Getting Right" post is now the vocabulary standard for enterprise AI conversations — use it to sharpen KwikGEO's pitch language. Also: add Duck.ai citation testing to the standard KwikGEO merchant audit. DuckDuckGo's +30% install surge means a real and growing share of search traffic routes through Claude and GPT-5 mini — merchants KwikGEO optimizes for major AI surfaces are already Duck.ai-compatible without extra work; surface this as a bonus data point in audit reports.