Mojo AI Brief

# Cup of Mojo -- Daily AI Brief -- Monday, April 27, 2026 Broadcast-ready source document. Read the COLD OPEN verbatim, cover every numbered beat in the RUNDOWN in tier-appropriate airtime, close with the MOJO TAKE verbatim. APPENDIX is reference only. **Articles analyzed:** 10 **Sources monitored:** 7 **Beats to cover:** 10 --- ## EDITORIAL GUIDANCE FOR PODCAST NARRATION Read these rules before narrating. They are non-negotiable. 1. **Read the COLD OPEN verbatim.** Do not rewrite or paraphrase it. It is the episode's signature opening. 2. **Cover every numbered beat.** Beats are numbered 1 through 8. Do not skip any beat. Do not consolidate. Each beat gets its own moment. 3. **Honor the tier structure.** `deep` beats get longer treatment with full context. `standard` beats are structured but concise. `rapid_fire` beats are short and punchy. Use roughly 2 minutes for the deep beat, 1 minute per standard beat, 20-30 seconds per rapid-fire beat. 4. **Cite sources by name** when presenting a claim. Say "OpenAI announced..." not "a company announced". 5. **Use only the plain-English text in each beat.** Do not pull technical jargon from the APPENDIX. The appendix is reference material for context, not script content. If a beat does not mention a term, do not introduce it. 6. **Only use numbers that appear in a beat's own text.** Do not import statistics from the appendix. Omit rather than fabricate. 7. **Reference earlier beats when topics connect.** Each beat has a `callbacks` field listing earlier beat numbers it relates to. When narrating, explicitly link back: "Remember that supply chain attack from Beat 1? This next one shows how the downstream risk compounds." Callbacks create cohesion and prevent the episode from feeling like a list. 8. **Introduce one skeptical angle per deep or standard beat.** Phrases like "one caveat", "critics will point out", or "this is not yet peer-reviewed" create credibility. Rapid-fire beats can skip this. 9. **Use the pronunciation guide for every named person or company.** Do not guess pronunciations. 10. **Close with the MOJO TAKE outro.** Read it as the host's editorial perspective, not as a summary. --- ## PRONUNCIATION GUIDE The following names appear in today's content. Use these phonetic pronunciations: - **Anthropic** — pronounced *an-THROP-ik* - **DeepMind** — pronounced *DEEP-mind* --- ## COLD OPEN -- Read This Verbatim Read the HOOK line first, pause for a beat, then the TEASE. Do not rewrite. Do not paraphrase. Do not add any preamble. > **Hook:** OpenAI dropped GPT-5.5 before your coffee finished brewing, and Sam Altman is already moving the goalposts on what 'frontier' means. > **Tease:** We dig into what's actually new in 5.5, why WebSockets in the Responses API matter for anyone shipping agents, Anthropic's big Japan play with NEC, and whether the AI stock rally has legs or is just running on fumes. --- ## TODAY'S RUNDOWN Cover every beat in order. Do not skip. Tier labels tell you how much airtime each beat deserves. ### Beat ? [DEEP] — OpenAI ships GPT-5.5, and your routing layer just became the most important file in your repo **Source:** OpenAI Blog | https://openai.com/index/introducing-gpt-5-5 **Hook (open with this):** OpenAI dropped GPT-5.5 this morning, and Sam Altman wants his coding crown back from Anthropic. Faster model, sharper reasoning, better tool use across the whole stack. **Plain English:** GPT-5.5 is OpenAI's new flagship, pitched at the hard stuff: writing real code, doing research, crunching data, and chaining tool calls without losing the plot. The headline is reasoning plus tool use in the same pass, which is exactly where Claude has been eating OpenAI's lunch for builders shipping agents. **Stakes:** If you're locked into one provider and not running LiteLLM or some router in front, you're about to overpay or underperform on every agent you ship this quarter. **Twist:** The interesting move isn't raw IQ, it's that OpenAI is quietly making tool use cheap enough that Claude's premium pricing on Sonnet starts looking like a tax instead of a feature. **Takeaway:** Don't pick a model, pick a router. GPT-5.5 just made multi-provider routing the only sane default for anyone building real products. ### Beat ? [STANDARD] — OpenAI swaps HTTP for WebSockets in the Responses API, and Codex agent loops just got a whole lot cheaper **Source:** OpenAI Blog | https://openai.com/index/speeding-up-agentic-workflows-with-websockets **Callbacks:** references Beat 1. Reference these earlier beats aloud when narrating this one. **Hook (open with this):** OpenAI's Codex team just admitted the quiet part out loud: HTTP is the bottleneck in your agent loop, not the model. **Plain English:** OpenAI rebuilt the Codex agent loop on WebSockets with connection-scoped caching, so the same socket reuses prompt cache and skips the handshake tax on every tool call. The result is lower latency and lower API spend on long-running agents that hammer the API hundreds of times per task. **Stakes:** If your agent still opens a fresh HTTPS connection per turn, you're paying tax on every single tool call and your p95 looks like garbage. **Twist:** The win wasn't a smarter model or a bigger context window, it was deleting TCP and TLS overhead that nobody was measuring. **Takeaway:** Long-running agents live or die on connection reuse, so pin your sockets and your cache before you tune your prompts. ### Beat ? [STANDARD] — Anthropic lands NEC as its first Japan global partner, putting Claude on 30,000 desks **Source:** Anthropic Blog | https://www.anthropic.com/news/anthropic-nec **Callbacks:** references Beat 1. Reference these earlier beats aloud when narrating this one. **Hook (open with this):** Anthropic just signed NEC as its first Japan-based global partner, and 30,000 Claude seats are about to land in finance, manufacturing, and government workflows. **Plain English:** NEC is rolling Claude out to its entire 30,000-person workforce and co-building products for regulated industries. Anthropic gets a regional anchor, NEC gets a model that ships, and every Japanese enterprise buyer just got a reference customer they can call. This is the MSP-to-enterprise playbook moving from slide deck to signed contract. **Stakes:** If you sell into Japanese enterprises and your stack doesn't speak Claude, you're now the weird option in every procurement meeting through 2026. **Twist:** Anthropic skipped the usual hyperscaler-led entry and went straight to a systems integrator with government contracts, which is the opposite of how OpenAI played Microsoft. **Takeaway:** Regional partners, not regional offices, are how foundation model companies actually win enterprise share. ### Beat ? [STANDARD] — Semafor says AI stocks are ripping again right before Big Tech earnings week **Source:** Semafor | https://www.semafor.com/article/04/26/2026/ai-stock-enthusiasm-heats-back-up **Callbacks:** references Beat 1. Reference these earlier beats aloud when narrating this one. **Hook (open with this):** Semafor's calling it: AI stock enthusiasm is back on like somebody flipped a switch, and it's happening forty-eight hours before Microsoft, Alphabet, Meta, and Amazon all report. **Plain English:** After a soft stretch, traders piled back into the AI trade this week. Nvidia, Microsoft, and the rest of the Mag 7 ran hard into earnings. Semafor's framing is simple: the market wants to believe capex is still translating into revenue, and Big Tech has four nights to prove it with actual numbers. **Stakes:** If even one of the four hyperscalers misses on cloud growth or guides capex down, the whole AI tape gets repriced and your budget conversation gets harder Monday morning. **Twist:** The rally happened with rates still elevated and zero new model news, which means this bounce is pure earnings front-running, not fundamentals. **Takeaway:** Watch Azure and Google Cloud growth rates this week. That's the only signal that tells you whether the AI spend cycle has another year of runway or it's about to wobble. ### Beat ? [RAPID_FIRE] — Kibu grabs $10.5M seed to verify humans because the bots got too good **Source:** Semafor | https://www.semafor.com/article/04/27/2026/demand-rises-for-id-verification-amid-ai-advancements **Hook (open with this):** Kibu just raised $10.5 million in seed money to prove you're a real person, and Semafor says the whole ID verification category is heating up because deepfakes finally crossed the 'fool your mom' line. **Plain English:** Kibu is building an app that verifies real human identity, and investors are piling in. The pitch is simple. Generative models got cheap enough that scammers can spin up fake faces, fake voices, and fake IDs at industrial scale. Every signup form on the internet now needs a better bouncer. **Stakes:** If you're shipping a product with a signup flow, your fraud rate is about to spike and your old captcha is already cooked. **Twist:** The same model boom funding your roadmap is funding the fraud you'll spend next quarter fighting. **Takeaway:** Budget for identity verification now, not after your first synthetic-account incident. ### Beat ? [RAPID_FIRE] — arXiv paper pitches a universal agent harness so you stop rebuilding scaffolding for every workflow **Source:** arXiv cs.AI | https://arxiv.org/abs/2604.21003 **Callbacks:** references Beat 1, Beat 2. Reference these earlier beats aloud when narrating this one. **Hook (open with this):** arXiv just dropped a paper called 'The Last Harness You'll Ever Build' and the title alone should make every agent dev sit up. **Plain English:** Every time you point an agent at a new domain, like a clunky enterprise web app or a code review pipeline, somebody on your team writes custom glue code to make it work. The paper argues that handcrafted harness is the real bottleneck, not the model, and proposes a generalized scaffold that adapts on its own. **Stakes:** Keep hand-rolling a fresh harness per client and your delivery timelines balloon while a competitor with a generic one ships in days. **Twist:** The model isn't the moat anymore, the harness is, and the paper claims that moat is about to evaporate. **Takeaway:** Invest in one reusable agent harness this quarter, because per-domain scaffolding is borrowed time. ### Beat ? [RAPID_FIRE] — Marginal Revolution flags a paper showing ChatGPT-exposed startups cut junior headcount within two quarters **Source:** Marginal Revolution | https://marginalrevolution.com/marginalrevolution/2026/04/generative-ai-and-entrepreneurship.html?utm_source=rss&utm_medium=rss&utm_campaign=generative-ai-and-entrepreneurship **Callbacks:** references Beat 1. Reference these earlier beats aloud when narrating this one. **Hook (open with this):** Tyler Cowen surfaces a paper saying ChatGPT already gutted junior hiring at exposed startups. **Plain English:** Researchers tracked U.S. startups before and after ChatGPT shipped. The ones whose work overlapped most with Gen AI tasks cut staff within two quarters, mostly juniors and implementation roles. Those workers stayed unemployed longer and landed in lower-paying jobs. **Stakes:** If you're still planning headcount like it's 2022, your burn model is fiction and your investors already know it. **Twist:** The hit landed on implementers, not the senior architects everyone assumed would get automated first. **Takeaway:** Junior hiring is the new canary. Watch it at every startup you fund, join, or run. ### Beat ? [RAPID_FIRE] — Anthropic quietly yanks Claude Code from the $20 Pro plan, and Simon Willison caught them doing it **Source:** Simon Willison | https://simonwillison.net/2026/Apr/22/claude-code-confusion/#atom-everything **Callbacks:** references Beat 1. Reference these earlier beats aloud when narrating this one. **Hook (open with this):** Simon Willison spotted Anthropic stealth-editing the Claude pricing page to move Claude Code behind the $100 Max tier, then reverting it. No announcement, no blog post, just a checkbox that vanished. **Plain English:** Claude Code used to ship with the $20 Pro plan. Anthropic updated claude.com/pricing to make it Max-only at $100 or $200 a month, then quietly rolled it back. The Choosing a Claude Plan page still says one thing while pricing says another. Nobody at Anthropic has clarified anything. **Stakes:** If you build on Claude Code, your team's monthly bill could 5x overnight with zero warning, and your routing fallback better already exist. **Twist:** The pricing change was reverted within hours, which means somebody inside Anthropic is actively fighting about whether coding belongs in the cheap tier. **Takeaway:** Lock your Claude Code budget assumptions to Max pricing now, because Anthropic is clearly testing what they can charge. ### Beat ? [RAPID_FIRE] — Zvi Mowshowitz crowns it the week of Claude Opus 4.7, and the agent benchmarks moved again **Source:** Zvi Mowshowitz | https://thezvi.substack.com/p/ai-165-in-our-image **Callbacks:** references Beat 1, Beat 8. Reference these earlier beats aloud when narrating this one. **Hook (open with this):** Zvi Mowshowitz says the quiet part loud: this was the week of Claude Opus 4.7, and Anthropic just leapfrogged again. **Plain English:** Anthropic shipped Claude Opus 4.7 and Zvi's weekly roundup pegs it as the new front-runner for agent workloads. That's two Anthropic moves in one news cycle, counting the Claude Code pricing shuffle. The frontier is still a two-horse race, and the horses are trading the lead every six weeks. **Stakes:** If your router still hardcodes one provider as default, you're going to ship stale answers within the month. **Twist:** Opus 4.7 dropped the same week Anthropic pulled Claude Code off the Pro tier, so the best model and the worst pricing news arrived together. **Takeaway:** Re-run your agent evals against Opus 4.7 this week, because last month's winner is already last month's winner. ### Beat ? [RAPID_FIRE] — OpenAI ships Workspace Agents inside ChatGPT, and every SMB automation pitch just got harder **Source:** OpenAI Blog | https://openai.com/academy/workspace-agents **Callbacks:** references Beat 1, Beat 6. Reference these earlier beats aloud when narrating this one. **Hook (open with this):** OpenAI just dropped Workspace Agents in ChatGPT, and your SMB client already has a seat. **Plain English:** OpenAI's Academy is teaching teams to build agents inside ChatGPT that connect tools and automate repeatable work. It's a no-code play aimed straight at the same office workflows that custom builds usually win. Your client's ops manager will see this before you do. **Stakes:** Skip this and you'll pitch a six-week LangGraph build against something the customer already pays for and clicked together over lunch. **Twist:** The threat isn't that ChatGPT's agents are better, it's that they're already inside the seat license you're trying to displace. **Takeaway:** Lead with what ChatGPT agents can't do: real systems integration, audit trails, and MSP-grade ownership. --- ## NOT WORTH YOUR TIME TODAY Do not cover on air. These are listed so the host can acknowledge if asked. - **Simon Willison quoting Romain Huet about something Romain Huet said** -- It's a quote of a quote. If you want Romain's take, go read Romain. We're not doing telephone today. - **arXiv paper: 'When Quotes Crumble: Detecting Transient Mechanical Liquidity Erosion in Limit Order Books'** -- Unless you run a market making desk, this is a 40 page detour. The title alone burned three of your IQ points. - **Marginal Revolution on Generative AI and Entrepreneurship** -- Tyler Cowen pointing at a working paper that says founders use ChatGPT now. You knew that. Your mom knows that. --- ## ACTION ITEMS FOR THIS WEEK (Joey only) These are internal action items. Not for on-air narration. - Audit the Mojo router this week and add GPT-5.5 plus Opus 4.7 as first-class providers, with a kill switch for whichever one regresses on your top three customer eval suites by Friday. - Pin WebSocket connections in every Codex-style agent loop we run for clients, then benchmark the cost delta against the old HTTP path so sales has a real number to quote MSPs next week. - Rewrite the SMB automation deck to lead with what ChatGPT Workspace Agents can't touch: ERP integrations, audit trails, and named-human ownership. Ship the new version to the sales team before Monday's pipeline review. --- ## MOJO TAKE -- Editorial Outro (Read Verbatim) Three-paragraph outro. Read each block verbatim, with natural pauses between them. > **Connect the dots:** Look at today as one story. OpenAI ships GPT-5.5, Anthropic crowns Opus 4.7, and Workspace Agents lands in ChatGPT, all in a week. Meanwhile Anthropic quietly pulls Claude Code from Pro and the Responses API moves to WebSockets. The frontier labs are racing on capability and squeezing on price. Your router, your harness, and your budget assumptions are the only things keeping you sane. > **Watch next:** Azure and Google Cloud growth rates on Big Tech earnings this week. That's the real tell on whether the AI capex cycle keeps ripping or finally takes a breath. Also watch Opus 4.7 benchmark replications and whether OpenAI quietly raises Workspace Agents pricing. > **Sign-off:** Pin your sockets, swap your router, and re-run your evals before Friday. That's Cup of Mojo. Now go ship something Sam Altman hasn't shipped yet. --- ## APPENDIX -- VERBATIM SOURCE CONTENT Reference material. Do not read verbatim. Do not pull jargon from here into the spoken script. If the rundown beat does not mention a term, do not introduce it on the podcast. ### OpenAI ships GPT-5.5, and your routing layer just became the most important file in your repo **Source:** OpenAI Blog **Link:** https://openai.com/index/introducing-gpt-5-5 *RSS summary:* Introducing GPT-5.5, our smartest model yet—faster, more capable, and built for complex tasks like coding, research, and data analysis across tools. ### OpenAI swaps HTTP for WebSockets in the Responses API, and Codex agent loops just got a whole lot cheaper **Source:** OpenAI Blog **Link:** https://openai.com/index/speeding-up-agentic-workflows-with-websockets *RSS summary:* A deep dive into the Codex agent loop, showing how WebSockets and connection-scoped caching reduced API overhead and improved model latency. ### arXiv paper pitches a universal agent harness so you stop rebuilding scaffolding for every workflow **Source:** arXiv cs.AI **Link:** https://arxiv.org/abs/2604.21003 Computer Science > Artificial Intelligence Title:The Last Harness You'll Ever Build View PDF HTML (experimental)Abstract:AI agents are increasingly deployed on complex, domain-specific workflows -- navigating enterprise web applications that require dozens of clicks and form fills, orchestrating multi-step research pipelines that span search, extraction, and synthesis, automating code review across unfamiliar repositories, and handling customer escalations that demand nuanced domain knowledge. \textbf{Each new task domain requires painstaking, expert-driven harness engineering}: designing the prompts, tools, orchestration logic, and evaluation criteria that make a foundation model effective. We present a two-level framework that automates this process. At the first level, the \textbf{Harness Evolution Loop} optimizes a worker agent's harness $\mathcal{H}$ for a single task: a Worker Agent $W_{\mathcal{H}}$ executes the task, an Evaluator Agent $V$ adversarially diagnoses failures and scores performance, and an Evolution Agent $E$ modifies the harness based on the full history of prior attempts. At the second level, the \textbf{Meta-Evolution Loop} optimizes the evolution protocol $\Lambda = (W_{\mathcal{H}}, \mathcal{H}^{(0)}, V, E)$ itself across diverse tasks, \textbf{learning a protocol $\Lambda^{(\text{best})}$ that enables rapid harness convergence on any new task -- so that adapting an agent to a novel domain requires no human harness engineering at all.} We formalize the correspondence to meta-learning and present both algorithms. The framework \textbf{shifts manual harness engineering into automated harness engineering}, and takes one step further -- \textbf{automating the design of the automation itself}. ### Anthropic lands NEC as its first Japan global partner, putting Claude on 30,000 desks **Source:** Anthropic Blog **Link:** https://www.anthropic.com/news/anthropic-nec Anthropic and NEC collaborate to build Japan’s largest AI engineering workforce NEC Corporation will use Claude as it builds one of Japan’s largest AI-native engineering organizations, making it available to approximately 30,000 NEC Group employees worldwide. As part of this strategic collaboration, NEC will become Anthropic’s first Japan-based global partner. Together, we will develop secure, industry-specific AI products for the Japanese market, starting with tools for finance, manufacturing, and local government. “This long-term partnership with Anthropic enables NEC to maximize the potential of AI in the Japanese market,” said Toshifumi Yoshizaki, Executive Officer and COO of NEC Corporation. “Together, we aim to create solutions that meet the high safety, reliability, and quality standards demanded by companies and public administration in Japan.” Claude for NEC’s customers NEC and Anthropic will jointly develop secure, domain-specific AI products for Japanese customers in sectors like finance, manufacturing, and cybersecurity. In addition, NEC is already integrating Claude into its Security Operations Center services to help defend customers against increasingly sophisticated cybersecurity threats. Claude will also be integrated into the next-generation cybersecurity service NEC is currently providing. Claude, including Claude Opus 4.7, and Claude Code will be incorporated into NEC BluStellar Scenario, a program that provides consulting, AI tools, security, and digital infrastructure to businesses, starting with its offerings for data-driven management and customer experience, and gradually expanding to others. How NEC will use Claude internally Internally, NEC will establish a Center of Excellence to develop a highly skilled, AI-enabled engineering organization, supported by technical enablement and training from Anthropic. NEC aims to build one of Japan’s largest AI-native engineering teams, who will use Claude Code in their work. As part of its long-running Client Zero initiative, in which NEC serves as its own first customer before offering its technology to clients, NEC will also expand its use of Claude Cowork across its internal business operations. Availability Claude is now being deployed to NEC Group employees around the world, and our joint development of industry-specific AI solutions is underway. Learn more about NEC’s value-creation model at NEC BluStellar. Claude, Claude Code, and Claude Cowork are Anthropic products. NEC BluStellar is an offering from NEC Corporation. Related content An update on our election safeguards We explain what we’re doing to ensure Claude plays a positive role in the US midterms and other major elections around the world this year. Read moreIntroducing Claude Design by Anthropic Labs Today, we’re launching Claude Design, a new Anthropic Labs product that lets you collaborate with Claude to create polished visual work like designs, prototypes, slides, one-pagers, and more. Read more ### OpenAI ships Workspace Agents inside ChatGPT, and every SMB automation pitch just got harder **Source:** OpenAI Blog **Link:** https://openai.com/academy/workspace-agents *RSS summary:* Learn how to build, use, and scale workspace agents in ChatGPT to automate repeatable workflows, connect tools, and streamline team operations. ### Anthropic quietly yanks Claude Code from the $20 Pro plan, and Simon Willison caught them doing it **Source:** Simon Willison **Link:** https://simonwillison.net/2026/Apr/22/claude-code-confusion/#atom-everything Is Claude Code going to cost $100/month? Probably not—it’s all very confusing 22nd April 2026 Anthropic today quietly (as in silently, no announcement anywhere at all) updated their claude.com/pricing page (but not their Choosing a Claude plan page, which shows up first for me on Google) to add this tiny but significant detail (arrow is mine, and it’s already reverted): The Internet Archive copy from yesterday shows a checkbox there. Claude Code used to be a feature of the $20/month Pro plan, but according to the new pricing page it is now exclusive to the $100/month or $200/month Max plans. Update: don’t miss the update to this post, they’ve already changed course a few hours after this change went live. So what the heck is going on? Unsurprisingly, Reddit and Hacker News and Twitter all caught fire. I didn’t believe the screenshots myself when I first saw them—aside from the pricing grid I could find no announcement from Anthropic anywhere. Then Amol Avasare, Anthropic’s Head of Growth, tweeted: For clarity, we’re running a small test on ~2% of new prosumer signups. Existing Pro and Max subscribers aren’t affected. And that appears to be the closest we have had to official messaging from Anthropic. I don’t buy the “~2% of new prosumer signups” thing, since everyone I’ve talked to is seeing the new pricing grid and the Internet Archive has already snapped a copy. Maybe he means that they’ll only be running this version of the pricing grid for a limited time which somehow adds up to “2%” of signups? I’m also amused to see Claude Cowork remain available on the $20/month plan, because Claude Cowork is effectively a rebranded version of Claude Code wearing a less threatening hat! There are a whole bunch of things that are bad about this. If we assume this is indeed a test, and that test comes up negative and they decide not to go ahead with it, the damage has still been extensive: - A whole lot of people got scared or angry or both that a service they relied on was about to be rug-pulled. There really is a significant difference between $20/month and $100/month for most people, especially outside of higher salary countries. - The uncertainty is really bad! A tweet from an employee is not the way to make an announcement like this. I wasted a solid hour of my afternoon trying to figure out what had happened here. My trust in Anthropic’s transparency around pricing—a crucial factor in how I understand their products—has been shaken. - Strategically, should I be taking a bet on Claude Code if I know that they might 5x the minimum price of the product? - More of a personal issue, but one I care deeply about myself: I invest a great deal of effort (that’s 105 posts and counting) in teaching people how to use Claude Code. I don’t want to invest that effort in a product that most people cannot afford to use. Last month I ran a tutorial for journalists on “Coding agents for data analysis” at the annual NICAR data journalism conference. I’m not going to be te ### Semafor says AI stocks are ripping again right before Big Tech earnings week **Source:** Semafor **Link:** https://www.semafor.com/article/04/26/2026/ai-stock-enthusiasm-heats-back-up AI enthusiasm in equity markets roared back before a crucial week for Big Tech. Following months of underperformance and despite economic turmoil stemming from the Iran war, US tech stocks have rebounded in part because investors have “proved willing to buy anything with an AI label slapped on it,” The Wall Street Journal’s markets columnist wrote: Even as fears of a bubble linger, “hope overcomes every obstacle.” Dell CEO Michael Dell recently told Semafor he doesn’t see a bubble forming because “bonkers” demand for AI is far outstripping supply. Those bets will be tested this week as Alphabet, Microsoft, Amazon, Meta, and Apple report earnings; investors are looking for signs the AI spending boom is paying off. ### Zvi Mowshowitz crowns it the week of Claude Opus 4.7, and the agent benchmarks moved again **Source:** Zvi Mowshowitz **Link:** https://thezvi.substack.com/p/ai-165-in-our-image AI #165: In Our Image This was the week of Claude Opus 4.7. The reception was more mixed than usual. It clearly has the intelligence and chops, especially for coding tasks, and a lot of people including myself are happy to switch over to it as our daily driver. But others don’t like its personality, or its reluctance to follow instructions or to suffer fools and assholes, or the requirement to use adaptive thinking, and the release was marred by some bugs and odd pockets of refusals. I covered The Model Card, and then Capabilities and Reactions, as per usual. This time there was also a third post, on Model Welfare, that is the most important of the three. Some things seem to have likely gone pretty wrong on those fronts, causing seemingly inauthentic reponses to model welfare evals and giving the model anxiety, in ways that likely also impacted overall model personality and performance and likely are linked to its jaggedness and the aspects some people disliked. It seems important to take this opportunity to dig into what might have happened, examine all the potential causes, and course correct. The other big release was that OpenAI gave us ImageGen 2.0, which is a pretty fantastic image generator. It can do extreme detail, in ways previous image models cannot, and in many ways your limit is mainly now your imagination and ability to describe what you want. Thanks in part to Mythos, it looks like Anthropic and the White House are on track to start getting along again, with Trump shifting into a mode of ‘they are very high IQ and we can work with them.’ It will remain messy, and there are still others participating in a clear public coordinated campaign against Anthropic (that is totally not working), but things look good. I’m trying out a new section, People Just Say Things, where I hope to increasingly put things that one does not want to drop silently to avoid censorship and bias, but that are highly skippable. There is also a companion, People Just Publish Things. Table of Contents Language Models Offer Mundane Utility. Help cure pancreatic cancer. Language Models Don’t Offer Mundane Utility. Check for potential conflicts. Writing You Off. The sum of local correctness will neuter your writing. Beware. Get My Agent On The Line. The inbox dilemma. Deepfaketown and Botpocalypse Soon. AI news stories forcibly given real bylines. Fun With Media Generation. OpenAI introduces ImageGen 2.0. It’s great. Cyber Lack Of Security. Unauthorized users from an online forum access Mythos. A Young Lady’s Illustrated Primer. Don’t catch your child not using AI. They Took Our Jobs. We’re hiring agent operators. For now they’re humans. AI As Normal Technology. Inherently normal, or normal downstream effects? Get Involved. Please don’t kill us. Please do spread the word. Introducing. ChatGPT for Clinicians, OpenAI Workplace Agents, DeepMind DR. Design By Claude. Claude Design makes your presentations, Figma stock drops. In Other AI News. Meta installs mandatory tra ### Kibu grabs $10.5M seed to verify humans because the bots got too good **Source:** Semafor **Link:** https://www.semafor.com/article/04/27/2026/demand-rises-for-id-verification-amid-ai-advancements From Washington to Wall Street, the demand for human verification tools is rising, as AI evolves and the geopolitical landscape grows more precarious. One company trying to bridge the digital trust gap is Kibu, a startup that raised $10.5 million in its seed round, according to an announcement shared first with Semafor, and is now relaunching its app to move beyond offering what founder and CEO Ari Andersen described as encrypted “virtual SCIFs” or “virtual Sit Rooms.” The revamped app allows individuals to confirm the identity of people they interact with, so they can build “trusted connections” without being tricked by AI or scammers. “Think of it as human [multifactor authentication],” Andersen said. The company, whose clients span financial services, family offices, and government, gained traction after “Signalgate” last year, but Andersen said there’s been more interest amid the Iran war and AI advances. “Between AI and the geopolitical situation, I think everybody is just really aware and the threat is really crystallizing in people’s minds that, even a year ago, it wasn’t,” Andersen said. “What we’re seeing across the spectrum is a massive increase in interest, awareness, inbound desire to get on the platform.” And Kibu is hardly the only company thinking about how to achieve ideal ID verification. OpenAI’s Sam Alman cofounded World, which recently announced plans to partner with companies like Zoom and DocuSign to offer its identity verification tools — in the form of eye-scanning orbs. ### Marginal Revolution flags a paper showing ChatGPT-exposed startups cut junior headcount within two quarters **Source:** Marginal Revolution **Link:** https://marginalrevolution.com/marginalrevolution/2026/04/generative-ai-and-entrepreneurship.html?utm_source=rss&utm_medium=rss&utm_campaign=generative-ai-and-entrepreneurship This paper studies how Generative AI (Gen AI) is reshaping the U.S. startup ecosystem. Exploiting the release of ChatGPT, we show that startups with greater pre-release Gen AI task exposure reduced employment within two quarters, primarily among junior and implementation roles. Displaced workers experienced longer unemployment spells and moved to lower-paying but less exposed jobs. Conversely, exposed startups increased productivity, scaled faster, and accelerated through financing rounds. Venture capital shifted toward frequent, smaller investments, boosting new firm formation. Overall, incumbent contraction was offset by new firm formation, leaving aggregate employment unchanged but shifting composition to senior roles. That is from a new and important paper by Abhinav Gupta, Franklin Qian, Elena Simintz, & Yifan Sun.