Mojo AI Brief

# Cup of Mojo -- Daily AI Brief -- Friday, April 24, 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:** 8 **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one skeptical angle per deep or standard beat.** Phrases like "one caveat", "critics will point out", or "this is not yet peer-reviewed" create credibility.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 quietly shipping WebSockets to make agents stop stalling mid-sentence. > **Tease:** Today: what GPT-5.5 actually changes, why the Responses API just got a speed injection, Anthropic planting a flag in Japan with NEC, and the mystery man writing checks to glimpse the future. Charts, chaos, and one ban on the phrase you know I hate. --- ## TODAY'S RUNDOWN Cover every beat in order. Do not skip. Tier labels tell you how much airtime each beat deserves. ### Beat ? [DEEP] — OpenAI drops GPT-5.5 and puts Claude's reasoning lead on notice **Source:** OpenAI Blog | https://openai.com/index/introducing-gpt-5-5 **Hook (open with this):** OpenAI just shipped GPT-5.5, and Sam Altman's crew is swinging straight at Anthropic's jaw. Faster, smarter, better at tool use, and aimed right at the workloads Claude's been quietly winning all year. **Plain English:** GPT-5.5 is OpenAI's new flagship, built for the hard stuff: coding, research, and chewing through data with tools in the loop. The headline number is reasoning, which is exactly where Claude has been eating OpenAI's lunch. And tool use, which is where every serious agent product lives or dies in 2025. **Stakes:** If you're routing traffic to Claude through LiteLLM right now, your cost and latency math just changed overnight and your CFO is going to notice. **Twist:** The surprise isn't that GPT-5.5 exists, it's that OpenAI shipped a reasoning upgrade before Anthropic did, flipping who's chasing who. **Takeaway:** Re-bench your model router this week. The Claude default you set six months ago is a bet, not a fact, and GPT-5.5 just moved the line. ### Beat ? [STANDARD] — OpenAI rewires Codex with WebSockets and cuts agent loop latency where it actually hurts **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 just shipped WebSockets in the Responses API, and Codex is the proof of life. **Plain English:** Every tool call in an agent loop used to pay the full HTTP tax: new connection, new auth, cold cache. OpenAI swapped that for a persistent WebSocket with connection-scoped caching, so the model keeps its context warm between turns. Codex got faster, and anyone running long agent loops on Responses gets the same free lunch. **Stakes:** If you're still polling a stateless endpoint in a ten-step agent, you're paying the handshake tax ten times while OpenAI's native stack pays it once. **Twist:** The win isn't a smarter model, it's a dumber pipe. Keeping the socket open did more for latency than half the prompt tricks people ship on Twitter. **Takeaway:** Connection scoping is the new batching. If your orchestrator can't hold state across tool calls, it's already losing to Codex on wall-clock time. ### Beat ? [STANDARD] — Anthropic plants its flag in Japan: NEC rolls Claude out to 30,000 employees and becomes co-dev partner **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 NEC employees are getting Claude on their desks. **Plain English:** NEC isn't just buying seats. They're co-building Claude-powered products for finance, manufacturing, and government clients across Japan. That's a Fortune 500-sized enterprise publicly standardizing on Claude, not hedging with a menu of models. Anthropic gets a distribution beachhead in a market where IBM, Microsoft, and the domestic telcos have owned enterprise IT for forty years. **Stakes:** If you're selling AI tooling into Japanese enterprises next year, NEC just became the reference architecture you're competing against or plugging into. **Twist:** The surprise isn't the seat count. It's that NEC, a company that usually builds its own stack, chose to co-develop on someone else's model instead. **Takeaway:** Enterprise standardization is happening, and right now Claude is the one getting standardized on. ### Beat ? [STANDARD] — Bill Nguyen pays six figures a month to run AI body doubles and says the chatbot era is already dead **Source:** Semafor | https://www.semafor.com/article/04/24/2026/the-man-who-is-paying-to-see-the-future **Callbacks:** references Beat 2. Reference these earlier beats aloud when narrating this one. **Hook (open with this):** Bill Nguyen, the guy who sold Lala to Apple and built Color, just told Semafor he's past chatbots. He's running AI body doubles that sit in meetings, draft deals, and act like a second him. **Plain English:** Nguyen isn't typing prompts anymore. He's spinning up persistent agents that know his context, his contacts, and his calendar, then letting them do the actual work while he reviews output. He's reportedly burning serious cash to live six months ahead of the rest of us. **Stakes:** If your competitor has a body double shipping work at 3 a.m. and you're still copy-pasting into a chat window, the gap compounds daily. **Twist:** The bottleneck isn't the model anymore. It's how much of your own context you're willing to hand over to one. **Takeaway:** Stop asking what the chatbot can do. Start asking what a version of you that never sleeps would do. ### Beat ? [RAPID_FIRE] — a16z's Charts of the Week: stablecoins quietly flip from wire-transfer hack to actual payment rail **Source:** a16z AI | https://www.a16z.news/p/charts-of-the-week-software-ate-the **Hook (open with this):** a16z just dropped Charts of the Week and buried the lede: stablecoin volume is shifting from transfers to real payments. **Plain English:** For years stablecoins were basically crypto plumbing, moving dollars between exchanges. a16z's new charts show that mix is flipping toward actual merchant and payroll payments. Meanwhile their 'Railroad GPT' chart argues AI is following the railroad buildout curve, and the promised productivity pop still hasn't shown up in the macro data. **Stakes:** Miss the stablecoin turn and you'll be the last shop wiring ACH while your competitor settles in USDC before lunch. **Twist:** Software ate the world, but the productivity gains everyone promised are still AWOL in the GDP numbers. **Takeaway:** Stablecoins are becoming a payments rail, not a trader toy. Price your next vendor contract accordingly. ### Beat ? [RAPID_FIRE] — arXiv paper flags silent output drift when you quantize LLMs down to int8 **Source:** arXiv cs.AI | https://arxiv.org/abs/2604.19790 **Hook (open with this):** arXiv researchers just put numbers on something your SRE has been whispering about: the int8 model on your edge box doesn't always agree with the bfloat16 model you tested. **Plain English:** Teams squeeze models into smaller number formats like int8 or float16 to run cheaper and faster on edge hardware. The paper shows those squeezed models quietly give different answers than the full-precision version, and standard evals miss it. No crash, no error, just a different answer. **Stakes:** If you evaluated in bfloat16 and shipped in int8, your production model is not the model you tested, and your regression suite won't catch it. **Twist:** The disagreements are small enough to slip past accuracy benchmarks but big enough to flip individual user outputs in production. **Takeaway:** Eval the exact precision you ship, on the exact hardware you ship it on, or you're flying blind. ### Beat ? [RAPID_FIRE] — FT poll of 4,000 workers: the top earners use AI daily, the kids don't **Source:** Marginal Revolution | https://marginalrevolution.com/marginalrevolution/2026/04/which-workers-are-using-ai-the-most-and-best.html?utm_source=rss&utm_medium=rss&utm_campaign=which-workers-are-using-ai-the-most-and-best **Callbacks:** references Beat 4. Reference these earlier beats aloud when narrating this one. **Hook (open with this):** The Financial Times polled 4,000 workers in the US and UK, and the adoption curve is not what your LinkedIn feed told you. **Plain English:** Over 60 percent of the highest earners use AI daily. Only 16 percent of lower earners do. And the youngest workers? Not the heaviest users. It's the well-paid mid-career crowd running circles around everyone else. **Stakes:** Price your product for Gen Z and you're chasing the wrong wallet. The buyers are senior, paid, and already hooked. **Twist:** Digital natives lost this race to forty-somethings with expense accounts and deadlines. **Takeaway:** Sell AI tools up-market, not down. The daily users have the budget and the pain. ### Beat ? [RAPID_FIRE] — GitHub Copilot pauses individual signups and walls Claude Opus 4.7 behind the $39 Pro+ tier **Source:** Simon Willison | https://simonwillison.net/2026/Apr/22/changes-to-github-copilot/#atom-everything **Callbacks:** references Beat 1, Beat 7. Reference these earlier beats aloud when narrating this one. **Hook (open with this):** GitHub just paused Copilot Individual signups, tightened usage caps, and locked Claude Opus 4.7 to the $39 Pro+ tier. Old Opus models? Gone. **Plain English:** Simon Willison flagged GitHub's quiet Copilot overhaul on the same day Anthropic flirted with a $100 Claude Code tier. Copilot is tightening limits, pausing new individual sign-ups, and pushing the good Claude model up a pricing rung. The unlimited coding assistant era is ending. **Stakes:** If your team is still on Copilot Individual, your renewal path may not exist and your best model just moved behind a paywall. **Twist:** GitHub pausing signups for its flagship consumer product means inference economics are breaking even for Microsoft. **Takeaway:** Coding assistant pricing is resetting upward. Budget for Pro+ tiers or plan your escape route now. ### Beat ? [RAPID_FIRE] — Zvi Mowshowitz calls it: the whole week belonged to Claude Opus 4.7 **Source:** Zvi Mowshowitz | https://thezvi.substack.com/p/ai-165-in-our-image **Callbacks:** references Beat 1, Beat 3, Beat 8. Reference these earlier beats aloud when narrating this one. **Hook (open with this):** Zvi Mowshowitz ran the tape on this week in AI and the verdict is one line: this was the week of Claude Opus 4.7. **Plain English:** Anthropic shipped Opus 4.7 and it ate the news cycle. Zvi's roundup, which tracks everything, basically says the other labs were playing for second place this week. That tracks with NEC standardizing on Claude and GitHub walling Opus 4.7 behind Pro+. **Stakes:** If your team's still treating Opus as the expensive backup model, you're reading the market a week late. **Twist:** GPT-5.5 launched the same week and still didn't own the conversation. **Takeaway:** When Zvi's weekly title is just your model name, you won the week. ### Beat ? [RAPID_FIRE] — OpenAI ships Workspace Agents and comes straight for your LangGraph stack **Source:** OpenAI Blog | https://openai.com/academy/workspace-agents **Callbacks:** references Beat 2, Beat 4. Reference these earlier beats aloud when narrating this one. **Hook (open with this):** OpenAI just dropped Workspace Agents inside ChatGPT, and the pitch is build, connect, scale without touching a framework. **Plain English:** OpenAI's new Academy track teaches teams to wire up multi-step agents in ChatGPT, hook into tools, and run repeatable workflows. No Python graph, no MCP server you babysit. Just ChatGPT swallowing the orchestration layer a lot of us built by hand. **Stakes:** If ChatGPT becomes the default place non-engineers build agents, your custom orchestration stack turns into a cost center nobody asked for. **Twist:** The real play isn't the agents, it's OpenAI running the training. Teach the workflow, own the runtime. **Takeaway:** Assume half your internal agent work migrates into ChatGPT by next year. Build the moat above it, not under it. --- ## NOT WORTH YOUR TIME TODAY Do not cover on air. These are listed so the host can acknowledge if asked. - **Simon Willison's pelican benchmark for GPT-5.5 via the semi-official Codex backdoor API** -- It's Simon drawing SVG birds through a side door. Cute hobby, not a signal you can trade on this week. - **Simon Willison's llm-openai-via-codex 0.1a0 plugin release** -- An 0.1 alpha plugin wrapping a backdoor API. Check back when it hits 1.0 or when OpenAI kills it, whichever comes first. - **Tyler Cowen's 'From the UAE' dispatch on Marginal Revolution** -- Travel notes, not tech news. Save it for the weekend coffee, not the Monday standup. --- ## ACTION ITEMS FOR THIS WEEK (Joey only) These are internal action items. Not for on-air narration. - Re-bench the Mojo router against GPT-5.5, Claude Opus 4.7, and whatever Gemini ships Friday. The Claude default is six months old. Run the eval on shipping precision, not fp16 in a notebook. - Rewrite the top-three customer pricing tiers around daily power users, not seat counts. The FT poll says the top earners are the ones actually paying. Price for them and stop discounting for tire-kickers. - Spin up a 90-minute war room on OpenAI Workspace Agents. Map every internal agent workflow that lives in LangGraph today and decide which ones become moat and which ones become ChatGPT tabs. --- ## MOJO TAKE -- Editorial Outro (Read Verbatim) Three-paragraph outro. Read each block verbatim, with natural pauses between them. > **Connect the dots:** Look at the through line: OpenAI ships GPT-5.5, rewires Codex, and drops Workspace Agents in the same week Zvi hands the crown to Claude Opus 4.7 and NEC standardizes 30,000 seats on Anthropic. The frontier is a two-horse race now, and every layer above the model, routers, orchestrators, coding tools, is getting repriced upward. Bill Nguyen is just skating to the puck. > **Watch next:** Watch whether Anthropic answers GPT-5.5 with an Opus 4.8 or Sonnet refresh, and watch GitHub Copilot's Pro+ conversion numbers. If Microsoft walls Claude behind $39 and nobody flinches, assistant pricing resets across the whole category. > **Sign-off:** That's your cup. Re-bench your router, price in the Pro+ hike, and go build something Bill Nguyen's body double can't. Joey out. --- ## 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 drops GPT-5.5 and puts Claude's reasoning lead on notice **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 rewires Codex with WebSockets and cuts agent loop latency where it actually hurts **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. ### Anthropic plants its flag in Japan: NEC rolls Claude out to 30,000 employees and becomes co-dev partner **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 ### arXiv paper flags silent output drift when you quantize LLMs down to int8 **Source:** arXiv cs.AI **Link:** https://arxiv.org/abs/2604.19790 Computer Science > Artificial Intelligence Title:Hidden Reliability Risks in Large Language Models: Systematic Identification of Precision-Induced Output Disagreements View PDF HTML (experimental)Abstract:Large language models (LLMs) are increasingly deployed under diverse numerical precision configurations, including standard floating-point formats (e.g., bfloat16 and float16) and quantized integer formats (e.g., int16 and int8), to meet efficiency and resource constraints. However, minor inconsistencies between LLMs of different precisions are difficult to detect and are often overlooked by existing evaluation methods. In this paper, we present PrecisionDiff, an automated differential testing framework for systematically detecting precision-induced behavioral disagreements in LLMs. PrecisionDiff generates precision-sensitive test inputs and performs cross-precision comparative analysis to uncover subtle divergences that remain hidden under conventional testing strategies. To demonstrate its practical significance, we instantiate PrecisionDiff on the alignment verification task, where precision-induced disagreements manifest as jailbreak divergence-inputs that are rejected under one precision may produce harmful responses under another. Experimental results show that such behavioral disagreements are widespread across multiple open-source aligned LLMs and precision settings, and that PrecisionDiff significantly outperforms vanilla testing methods in detecting these issues. Our work enables automated precision-sensitive test generation, facilitating effective pre-deployment evaluation and improving precision robustness during training. ### OpenAI ships Workspace Agents and comes straight for your LangGraph stack **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. ### Bill Nguyen pays six figures a month to run AI body doubles and says the chatbot era is already dead **Source:** Semafor **Link:** https://www.semafor.com/article/04/24/2026/the-man-who-is-paying-to-see-the-future The News On a recent morning at his Atlanta office, Bill Nguyen tore open a large package and lifted out a high-end desktop computer. “This thing weighs like 30 pounds!” said Nguyen, a serial entrepreneur whose boundless energy and wiry frame make his age — 55 — slightly difficult to fathom. When he peered inside the box, he unearthed a top-of-the-line Nvidia RTX 5090 graphics card. “I suspect it is going to make me train something.” Nguyen didn’t order the machine, but he had a good idea of who did — his AI assistant. It’s another component in his bid to build an AI system that can operate as his proxy — moving far beyond a chatbot to create a virtual body double that can absorb enough information about how he communicates and what he values, so that it can make decisions and take actions on his behalf. Nguyen, who has made a small fortune selling multiple companies to Apple and recently launched a voice-recognition startup called Olive, said his personal AI has all but taken over his life. Now when he wakes up most mornings, he consults the agenda his AI assistant has crafted for him, and then spends his days following its directions. The AI has permission to email people on his behalf, and sometimes sets up in-person meetings with people he has never met. It listens to conversations he has with his three kids, and then suggests parenting advice, which he says has improved his relationship with them. Nguyen is a portrait of an emerging class of token-maxxing power users who are plunging tens of thousands of dollars to MacGyver next-level AI assistants, not by waiting for the next big model release, but by orchestrating today’s models in loops, with more computing power, more passes, and more automated checking — and a massive dose of risk tolerance. The idea is to give the system an unlimited amount of tokens and access to every conceivable piece of relevant data. “I didn’t ask it to help me,” Nguyen said. “I asked it to be me.” Know More If Nguyen’s account sounds out of reach for most people, that’s because, for now, it is — at least financially. Nguyen declined to say publicly how much he is spending, but the higher levels of reliability and capability from the AI assistant come, in part, by spending gobs of money on more tokens, the small units of text that AI providers bill for when developers use their models through APIs. While many people interact with AI through $20 to $200 per-month subscriptions, Nguyen said he pays per token and runs multiple models repeatedly, sometimes in parallel. At first, there was a sticker shock. “I’m like, oh my God, this is really expensive.” It would be unaffordable for most people, he said. While Nguyen is spending time with his family or out running his business, in the background there’s a kind of endless conversation with a bot that’s sending into motion many chatbots from several different providers all working in unison. He calls it “agentic scaling,” industry shorthand for adding new capabilities by ### GitHub Copilot pauses individual signups and walls Claude Opus 4.7 behind the $39 Pro+ tier **Source:** Simon Willison **Link:** https://simonwillison.net/2026/Apr/22/changes-to-github-copilot/#atom-everything 22nd April 2026 - Link Blog Changes to GitHub Copilot Individual plans (via) On the same day as Claude Code's temporary will-they-won't-they $100/month kerfuffle (for the moment, they won't), here's the latest on GitHub Copilot pricing. Unlike Anthropic, GitHub put up an official announcement about their changes, which include tightening usage limits, pausing signups for individual plans (!), restricting Claude Opus 4.7 to the more expensive $39/month "Pro+" plan, and dropping the previous Opus models entirely. The key paragraph: Agentic workflows have fundamentally changed Copilot’s compute demands. Long-running, parallelized sessions now regularly consume far more resources than the original plan structure was built to support. As Copilot’s agentic capabilities have expanded rapidly, agents are doing more work, and more customers are hitting usage limits designed to maintain service reliability. It's easy to forget that just six months ago heavy LLM users were burning an order of magnitude less tokens. Coding agents consume a lot of compute. Copilot was also unique (I believe) among agents in charging per-request, not per-token. (Correction: Windsurf also operated a credit system like this which they abandoned last month.) This means that single agentic requests which burn more tokens cut directly into their margins. The most recent pricing scheme addresses that with token-based usage limits on a per-session and weekly basis. My one problem with this announcement is that it doesn't clearly clarify which product called "GitHub Copilot" is affected by these changes. Last month in How many products does Microsoft have named 'Copilot'? I mapped every one Tey Bannerman identified 75 products that share the Copilot brand, 15 of which have "GitHub Copilot" in the title. Judging by the linked GitHub Copilot plans page this covers Copilot CLI, Copilot cloud agent and code review (features on GitHub.com itself), and the Copilot IDE features available in VS Code, Zed, JetBrains and more. ### Zvi Mowshowitz calls it: the whole week belonged to Claude Opus 4.7 **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 ### a16z's Charts of the Week: stablecoins quietly flip from wire-transfer hack to actual payment rail **Source:** a16z AI **Link:** https://www.a16z.news/p/charts-of-the-week-software-ate-the Charts of the Week: Software Ate the World Railroad GPT; Stablecoins volumes are shifting from transfers to payments; The Next Decade of News; See ya later, productivity gains America | Tech | Opinion | Culture | Charts We’re excited to welcome Lisha Li to the a16z Infra team. See her announcement here. -AD Software ate the world Obviously, we’re biased, but it’s hard to overestimate just how important technology is to the global economy. You might even say that software, literally, ate the world: The top 10 public companies by market cap are larger than the combined GDPs of the G7 (ex-US)--and that would be true, even if one excluded Saudi Aramco, which no one would consider a “tech” company. (Although it was founded in San Francisco!) To be fair, the Top 10 list is more “tech and semis [and however one would categorize Tesla and Apple]” than pure-play software, but the point stands: tech isn’t just a big deal, it’s the biggest deal. And tech’s global takeover has all happened fairly recently: The top 10 techcos were a small fraction of the G7 (ex-US), until cloud really began to hit its stride in ~16-’17. From that point, it took less than a decade for their combined market cap to eclipse the rest-of-world’s GDP (ex-China). Tech’s ascendancy isn’t just a changing of the guard, either. The biggest companies are much bigger than they were, even just 10 years ago: The combined market cap for the 10 largest companies in the S&P is ~6x larger than it was in 2015, and comprises ~2x larger share of the total index. To be sure, there was in fact a changing of the guard. The composition of the Top 10 changed-over dramatically, relative to prior decades. By 2025, there were only three holdovers from the previous decade, and only one (Microsoft, a tech company), from the decade before that. If you were an investor back in 2015, and you were trying to model comparable outcomes for techcos based on the biggest companies in the index, you would have undercounted the upside by a country mile (or 6). Fundamentally, tech “busted the model,” by redefining the outer limit of how large companies could become. And the outer limit still appears to be moving outwards! Indeed, tech has become even more central to the global growth story, as of late. Last week, we showed that Tech earnings are expected to grow ~2x faster than the rest of the market. But, if you look back even further, you would notice that tech is contributing an historically large share of the market’s overall earnings growth: Since 2023, Tech has been responsible for ~60%+ of earnings growth (give-or-take), market-wide. Other than a brief moment for energy in the early aughts, no other sector has played such a central role in the earnings story (and for quite so long) this century. At this point, it’s fair to say that tech isn’t just a cycle, it is the cycle. Railroad GPT We just told you that tech is an unprecedentedly large deal, but that’s not actually true. In the industrial era, no sector has ev ### FT poll of 4,000 workers: the top earners use AI daily, the kids don't **Source:** Marginal Revolution **Link:** https://marginalrevolution.com/marginalrevolution/2026/04/which-workers-are-using-ai-the-most-and-best.html?utm_source=rss&utm_medium=rss&utm_campaign=which-workers-are-using-ai-the-most-and-best An FT poll of 4,000 workers in the US and UK shows adoption is heavily skewed towards the best-paid workers: more than 60 per cent use AI daily, compared with just 16 per cent of the lower earners. Link here. Note also that the youngest workers are not those who use AI the most, rather it is workers in their 30s. Men in the workplace are using AI more than women are. A very good piece by Madhumita Murgia and John Burn-Murdoch.