> daily_signal(2026_05_18)

The class of 2026 booed Eric Schmidt every time he said AI. The people building it didn't slow down.

PickBits Daily Signal · Monday, May 18, 2026

By Mark Pickering · 6 min read · May 18, 2026

// tl;dr

At University of Arizona's Saturday commencement, graduates booed former Google CEO Eric Schmidt repeatedly when he talked about AI. A day earlier at UCF in Orlando, a Tavistock Development Company executive was booed for calling AI "the next industrial revolution." NBC News, TechCrunch, and Cybernews all framed it the same way: the class entering the job market is loud about who is benefiting from the technology they are being told to celebrate.
Verizon became the first telecom carrier to join Anthropic's Project Glasswing on Friday, May 15. The carrier will use Claude Mythos Preview, the same offensive cyber model Anthropic has refused to release publicly, to probe its own network for vulnerabilities. AT&T told Mobile World Live it is also in the coalition. The model whose capability the EU has been arguing about now sits inside two US phone networks.
OpenAI consolidated ChatGPT, Codex, and the developer API under co-founder Greg Brockman on Friday, four days before Google I/O. Brockman permanently takes product strategy; the three teams collapse into one organization aimed at shipping the desktop super-app (ChatGPT + Codex + the Atlas browser) faster. Fidji Simo, OpenAI's CEO of Applications, hired in May 2025 from Instacart, runs the consumer surface.
General Motors published the AI skills profile for the IT roles it is hiring to replace the roughly 600 IT workers it laid off this month. TechCrunch's Sunday report named the seven categories: AI-native development, data engineering, cloud-engineering, agent and model development, prompt engineering, AI-workflow design, and analytics. It is the cleanest public blueprint yet of how a Fortune 50 is reshaping IT around AI tooling rather than around the people who used to run it.

Saturday at the University of Arizona, the class of 2026 booed Eric Schmidt every time he opened his mouth about AI. They booed when he likened the moment to the arrival of the personal computer. They booed when he said the future of AI was "unwritten" and the graduates "have real power to shape it." The pattern was not isolated: at UCF the day before, a development-company executive was booed for calling AI "the next industrial revolution." Meanwhile the companies giving them that future spent the same weekend tightening their grip. Verizon let Anthropic's most restricted model start probing its phone network. OpenAI collapsed three of its product teams under one executive to ship a desktop super-app before Google I/O. GM published the skills profile for the AI-fluent IT workforce that is going to replace the workforce it just cut. Today's signal sits in the gap between what those companies shipped this weekend and what the people graduating into the economy they are reshaping said about it.

Arizona graduates booed Eric Schmidt's AI lines on Saturday. Verizon let an experimental Anthropic model probe its phone network. OpenAI put one person in charge of ChatGPT, Codex, and the API four days before Google I/O. GM published the AI skills profile for backfilling the IT cuts.

1. Saturday at Arizona, the class of 2026 booed Eric Schmidt every time he said AI.

At the University of Arizona's 162nd commencement on Saturday, May 16, former Google CEO Eric Schmidt was booed repeatedly while discussing artificial intelligence, with the booing intensifying every time the topic returned to AI and data centers. Schmidt was wearing the academic regalia given to him in 2014 alongside an honorary doctorate. He told the roughly 10,000 graduates that they were "entering an era of profound transformation," likening AI to the personal computer. He acknowledged in the speech: "We thought we were adding stones to a cathedral of knowledge that humanity had been constructing for centuries, but the world we built turned out to be more complicated than we anticipated." The boos got louder. Student groups (FORCE, Students for Socialism, the Women and Gender Resource Center, and Pride Alliance) had circulated a flyer ahead of the ceremony asking graduates to turn their backs or boo during Schmidt's speech, citing both his AI track record and a separate civil suit filed by a former partner.

The day before, at the University of Central Florida commencement in Orlando, an executive from Tavistock Development Company drew boos for telling the audience that "the rise of artificial intelligence is the next industrial revolution." TechCrunch's framing on Sunday was direct: if you are giving a commencement speech in 2026, do not mention AI. The class of 2026 is the first graduating cohort entering a labor market where the major reason employers cite for entry-level cuts is the same technology their commencement speakers are flying in to celebrate. As of last week, US tech layoffs in 2026 had passed 113,000 workers across roughly 180 events, with Meta's 8,000-job cut kicking in this Wednesday and GM's IT cut last week (covered in Story 4) the most recent named example. The booing is a leading indicator, not a one-off: the pitch ("AI as cathedral, you as builders") stopped landing with the people the technology is being deployed against.

NBC News coverage of Eric Schmidt being booed at University of Arizona commencement for AI remarks — nbcnews.com · May 17, 2026

Why this matters: If you have a kid graduating into the workforce this month, or you are still on the job hunt after a layoff, the most-cited reason employers give for entry-level cuts is the same technology your school is paying its commencement speakers to celebrate. The class of 2026 is the first one loud enough about it to boo on camera. Action this week: if you are a parent of a 2026 graduate, watch the actual speech on YouTube before reading the takes (NBC News uploaded the full Schmidt segment) so you can see when the boos hit and why; the audience reaction tracks one specific frame, not the speaker. If you are a graduate or a recent layoff who works in tech, open layoffs.fyi and trueup.io and write down which companies in your field have cut staff in the last 60 days while increasing AI spending; that list is the place to focus your applications away from. If you sit on a hiring panel for entry-level roles right now, write a one-line answer to the question your interviewees are about to ask: are you replacing a person who left or a job that AI replaced? You will be asked it this summer either way.

nbcnews.com/tech/tech-news/former-google-ceo-booed-graduation-speech-ai

2. Your phone carrier just let an experimental Anthropic model start probing its network.

Telecom security. On Friday, May 15, Verizon announced that it has joined Project Glasswing, the cross-industry coalition Anthropic launched in April 2026 to give a small set of organizations access to Claude Mythos Preview for defensive cybersecurity work. Verizon is the first US telecommunications carrier in the group; AT&T told Mobile World Live the same week that it is also participating. The model Verizon is now running across its network is the same one Anthropic has refused to release publicly because of its offensive capability: in earlier testing, Mythos identified a 27-year-old vulnerability in OpenBSD (a security-focused operating system used in firewalls and routers) and a 16-year-old bug in FFmpeg that automated scanners had passed over five million times.

The Project Glasswing coalition launched on April 7, 2026 with AWS, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks as the founding members. Anthropic committed up to $100 million in Mythos usage credits for participants and $4 million in direct donations to open-source security organizations. The addition of telecoms expands the surface area materially: phone networks carry the data plane for every authentication factor used by every other Glasswing participant. Verizon's announcement framed it as the carrier "using AI that finds decades-old security bugs"; the practical read is that the same model the European Commission has been formally arguing for access to, the one a Palo Alto security firm used to bypass Apple's M5 memory-integrity defense in five days (covered in Friday's daily), is now inside a network roughly 145 million wireless customers route their calls and text messages through. The defensive posture is the right one; the access-control question is now bigger.

Verizon press release announcing it joined Anthropic's Project Glasswing as first telecom — verizon.com · May 15, 2026

Why this matters: If Verizon (or AT&T) is your phone carrier, an experimental Anthropic model whose offensive capability was deemed too dangerous to release is now scanning the network your two-factor authentication codes ride on. The intent is defensive; the access control is what to watch. Action this week: if your most important accounts (email, bank, work SSO) use SMS as a recovery factor, switch them to an authenticator app or a hardware key this week (Google Authenticator, 1Password, Aegis, YubiKey). The Verizon-Glasswing news does not require this, but it is a clean reminder that "your phone carrier has good intentions" is not the same as "your phone carrier cannot be compromised." If you run IT or security at a small business, open your carrier's enterprise security portal and confirm which SS7 / SIM-swap protections are turned on for executive lines; that is the surface a successful Mythos-class finding on the network would expose first. If you are a journalist or a public official with a real threat model, ask your carrier directly (in writing) whether your account is enrolled in their highest-tier port-out protection program.

verizon.com/about/news/verizon-joins-anthropics-project-glasswing

If a friend forwarded this to you and you want the signal in your inbox five mornings a week, subscribe at pickbitsai.substack.com. The daily is free.

3. Four days before Google I/O, OpenAI put one person in charge of ChatGPT, Codex, and the API.

OpenAI reorg before Google I/O. On Friday, May 15, OpenAI told employees that co-founder and president Greg Brockman will permanently lead all product strategy, merging the ChatGPT, Codex, and developer API teams into a single organization. The reshuffle lands four days before the Google I/O keynote on Tuesday, May 19, where Google is expected to unveil a new Gemini model (Gemini 3.2 Flash leaked in the iOS app and AI Studio on May 5) along with the consumer-laptop OS (Aluminium OS) and the Android XR glasses launch. Fidji Simo, OpenAI's CEO of Applications hired from Instacart in May 2025, continues to run the consumer Applications group. The internal frame, per a Wall Street Journal report on March 19 and a Tech Times follow-up Friday: shipping the previously announced desktop super-app (ChatGPT + Codex + the Atlas browser) was being slowed down because three product teams owned three pieces of one user experience, and the quality bar Simo wanted to hit required one team.

The Atlas piece is the part that matters for everyone else. Atlas is the AI-first web browser OpenAI shipped earlier this year; folding it into the same product organization as ChatGPT and Codex is the structural prerequisite to the "always-on AI layer" device-and-OS strategy Bloomberg reported on May 14 ("AI-first device" in development, partnerships with MediaTek and Qualcomm for chip supply). The four-day distance from Google I/O is not a coincidence: Google's keynote will showcase Gemini-everywhere, and a fragmented ChatGPT + Codex + Atlas product surface would have read as the slower one. The reorg also pushes OpenAI further into Apple's hardware footprint at the same time OpenAI is preparing a breach-of-contract notice against Apple over the Siri-ChatGPT integration (covered in Sunday's daily). The model layer is consolidating in two directions at once: Google ships one app/model/OS stack tomorrow; OpenAI just made the org chart match.

Tech Times coverage of OpenAI merging ChatGPT Codex API under Brockman before Google I/O — techtimes.com · May 16, 2026

Why this matters: If you use ChatGPT, Codex, or the Atlas browser today, the next version is going to feel like one product instead of three, and Google's competing version of the same idea ships on Tuesday. The race you are watching from the outside is who can get the AI layer inside everyone's daily computing first. Action this week: Go to io.google Tuesday at 10am PT and watch the keynote live for Gemini-everywhere plumbing announcements. Then open the ChatGPT changelog at chat.openai.com/changelog and the Atlas release notes Friday and again the following Friday to see what OpenAI shipped in response. If you build internal tools on the OpenAI developer API, open your integration docs right now and write down which endpoints are on your critical path and what your sunset risk looks like if the API team shifts to consumer cadence. If you are an enterprise architect with a Microsoft 365 Copilot rollout and a parallel ChatGPT Enterprise pilot, pull up your existing contract terms on data residency and API stability before your next renewal - the Apple-OpenAI dispute is the first concrete data point on how OpenAI treats partners when growth misses internal targets.

techtimes.com/articles/316730/20260516/openai-unifies-chatgpt-codex-developer-api-under-co-founder-brockman-four-days-before-google-i-o

4. GM published the AI skills profile for the IT roles replacing the workers it just cut.

Continuing the GM IT layoff arc from 5/12. On Sunday, May 17, TechCrunch published a follow-up to its earlier reporting on General Motors' May layoff of roughly 600 salaried IT workers (about 10 percent of its IT department). The new piece names the seven categories of skills GM is hiring for to backfill those cuts: AI-native development, data engineering and analytics, cloud-based engineering, agent and model development, prompt engineering, AI-workflow design, and AI-fluent product management. The framing from inside GM is "skills swap, not headcount reduction," but TechCrunch's read is more direct: this is the cleanest public blueprint of how a Fortune 50 is reshaping its IT workforce around AI tooling rather than around the people who used to run it. The roles GM is closing are systems-of-record IT (ServiceNow operators, legacy application support, traditional DBA work); the roles it is opening price the same labor inputs (operations, data, system stitching) inside the AI tool layer, not outside it.

Three pieces of context for the line workers and IT managers reading this. First, GM is not unique. Cisco's 4,000-person cut last Wednesday (covered 5/14) was the same shape: a quarterly record-revenue print, a stock pop, and an explicit reshape of who reports to whom. Walmart's 1,000 corporate cuts to fund "super agents" (covered 5/15) was the same. Second, the skills GM named are not new disciplines; they are AI-flavored relabels of disciplines that already exist inside any working IT shop. "Agent and model development" is what platform engineering is becoming; "prompt engineering" is what API contract design looks like in a model-first world; "AI-workflow design" is what BPM and integration architecture become when one of the nodes is a language model. Third, the skills list maps almost exactly onto the existing PickBits University tracks for Vibe Coder, Vibe Management, and Field Guides; the gap between "the worker who got laid off" and "the worker GM is hiring next" is closer than the press release implies, and the people who have already started building on the new stack are the ones who close it.

techcrunch.com · May 17, 2026

Why this matters: If you work in IT at any large company, the GM skills list is the playbook your CIO is about to copy-paste into next quarter's hiring requisition. The seven categories tell you which of your current responsibilities will be relabeled and which will be dropped, and the timing tells you the window to learn the relabeled versions inside your current employer before you are asked to interview for them somewhere else. Action this week: open layoffs.fyi for the IT/operations category and find the 10 most recent cuts at companies in your industry, then cross-reference the AI-fluent IT job postings at the same companies (LinkedIn, Indeed, the company careers page); the delta between "what was cut" and "what is being hired" is the specific skills swap you are competing on. If you are a Business Analyst, pick one of the agent-development frameworks GM named (LangGraph, OpenAI Agents SDK, Claude Code) and ship one working agent against an internal workflow you already understand this week, end to end (define the trigger, write the prompt, wire one tool call, validate the output) - the agent does not have to be good; it has to exist. If you are a DBA or platform engineer, the prompt-engineering relabel is your strongest move; spend two hours rewriting one of your existing change-management runbooks as a structured agent prompt and test it against last week's actual change tickets. If you are an IT manager whose team is up for review, the conversation with your CIO is not "do we use AI"; it is "which two of the seven skills do we re-skill into this quarter and which do we hire for." Pick before you are picked.

techcrunch.com/2026/05/17/techcrunch-mobility-the-ai-skills-arms-race-is-coming-for-automotive

» What to watch this week

Google I/O keynote, Tuesday May 19, 10am PT. A new Gemini model (Gemini 3.2 Flash or 3.5), the consumer-laptop OS code-named Aluminium, Android XR glasses, and Android 17 are the four pieces of the keynote that determine whether the OpenAI-Brockman consolidation read above (Story 3) gets a counter-narrative or a confirmation. The Gemini Omni video-generation product surfaced in the Gemini app UI on May 2 is the most-leaked unannounced piece; whether Google formally names it Tuesday is the indicator on cross-modal generation strategy.
Meta Tuesday May 20 cut effective date. The 8,000-person reduction kicks in 24 hours after Google I/O. The number to watch is the unit distribution: Reality Labs, recruiting/HR, ads-and-social engineering, and global ops. The mix decides whether Meta is reshaping (Reality Labs absorbs most) or doing AI-attributed cost-cutting (cuts pull from core ads).
Connecticut SB 5 signing. The omnibus AI and online-safety bill cleared both chambers May 1; Governor Lamont's spokesperson said he "looks forward to signing" but the formal signature has not landed yet. The signing statement on the AI Responsibility and Transparency Act sections (employment-decision disclosure, AI-chatbot safety protocols, synthetic-content provenance) is the marker for how strong CT's enforcement posture will be in October when the law takes effect.
Normal, Illinois Monday May 18 council vote on a six-month data-center moratorium. Calipatria, California's first public hearing on its own freeze. Charlotte, North Carolina's 150-day pause vote in the same window. Three more local decisions in 10 days on top of last week's PJM-monitor finding and the 71 percent Gallup poll.
Texas AG Paxton's opinion on county data-center moratoriums. State Sen. Paul Bettencourt asked Paxton to investigate Hill County's authority to pass its May 12 one-year pause; a state-preemption opinion (or a Provident Data Centers lawsuit against Hill County) is the next milestone. Provident's 300-acre Hillsboro project is the named test case.
First named telecom Glasswing finding. Verizon's announcement said it will "share findings and best practices across industries." The first CVE that lists Mythos Preview as the discovery engine on telecom infrastructure (SS7, signaling, network management) is the practical confirmation the model is generating useful outputs at this scale; absence of that disclosure 60-90 days from now is its own signal.

Tomorrow's signal lands here.