Next Week in The Sequence:

We continue our series about model distillation.
In the AI of the Week, we are going to discuss OpenAI’s recent analysis of coding benchamarks.
In the opinion section, we are going to debate Meta’s opportunities and tremendous challenges to catch up with the AI frontier labs.

Subscribe and don’t miss out:

TheSequence is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Editorial: GPT-5.6, Grok 4.5, Muse Spark 1.1 and the Post-Chatbot Stack

Frontier AI labs have settled into a near-monthly release cadence: a new model, agent, or interface makes last quarter’s frontier feel like legacy infrastructure. This week’s GPT-5.6, GPT-Live, ChatGPT Work, Grok 4.5, and Muse Spark 1.1 reveal a shift. The model is becoming a runtime, the chat window a control plane, and “answering” is giving way to execution.

GPT-5.6 makes that transition explicit. OpenAI split the family into Sol, Terra, and Luna, optimizing for intelligence and performance per dollar. The geekiest feature is programmatic tool calling, which lets the model write programs to coordinate tools and process intermediate results. Add parallel subagents, and inference starts looking less like autocomplete and more like distributed systems engineering.

GPT-Live attacks another bottleneck: the turn-based interface. Its full-duplex architecture can listen and speak simultaneously, decide when to interrupt or remain silent, and delegate deeper reasoning while keeping the conversation alive. This is not merely better text-to-speech. It is an event loop for human-machine collaboration, replacing “your turn, my turn” with something closer to shared cognitive bandwidth.

ChatGPT Work completes the stack. It can operate across connected apps, websites, and files; stay on a project for hours; and produce editable documents, spreadsheets, presentations, and sites. The product transition is subtle but massive: the unit of value is no longer a response. It is a finished artifact—and, increasingly, an ongoing process.

Meta’s Muse Spark 1.1 makes the race more crowded and cheaper. It combines a million-token context window with multimodal perception, coding, computer use, and multi-agent orchestration. Its most interesting trick may be active context management: compacting extended sessions without losing the state needed later, while choosing between scripting an action and manipulating an interface directly. The paid Meta Model API also marks a strategic turn. Meta is not just releasing models; it wants to sell metered intelligence.

Grok 4.5 arrives at similar coordinates from another direction. Built for coding, agentic tasks, and knowledge work, it pushes into application generation and complex productivity artifacts. Its aggressive pricing adds pressure to the execution layer. Model competition is becoming a race to deliver the cheapest reliable unit of completed work.

The timing is revealing. Frontier labs are vertically integrating models, voice interfaces, agents, browsers, desktop environments, and artifact layers. They are fighting for ownership of the loop between intent and outcome. Whoever owns that loop gets the feedback data, developer ecosystem, and switching costs.

Of course, the abstraction layer brings new failure modes. Long-running agents need permissions, audit trails, checkpoints, and graceful rollback. A hallucinated paragraph is annoying; a hallucinated workflow touching your CRM, filesystem, or financial model is an incident.

That is why this week matters. The frontier is shifting from raw IQ to systems design: orchestration, latency, token efficiency, computer use, memory, governance, and interface. The winner may not top every static benchmark. It may be the model that best schedules intelligence across tools, time, and humans.

The chatbot era is not ending. It is being compiled into infrastructure.

AI Research

Separating signal from noise in coding evaluations

AI Lab:OpenAI

Summary:An audit of the popular SWE-Bench Pro coding benchmark reveals that approximately 30% of its tasks are broken due to issues like overly strict tests, underspecified prompts, and misleading instructions. As these flaws misrepresent true model capabilities, OpenAI retracts its recommendation for the benchmark and emphasizes the need for new, rigorously designed evaluations built by experienced software developers.

An off switch for dual-use knowledge in AI models

AI Lab:Anthropic (in collaboration with AE Studio)

Summary:This research introduces Gradient-Routed Auxiliary Modules (GRAM), a method that compartmentalizes specific categories of dual-use knowledge—such as virology or cybersecurity—into dedicated, removable neural network modules during training. By using GRAM, developers can effectively toggle dangerous capabilities on or off for different deployment environments without having to expensively retrain the entire model or degrade its general performance.

SkillOpt-Lite: Better and Faster Agent Self-evolution via One Line of Vibe

AI Lab:LMMS-Lab, NTU MMLab, and Microsoft

Summary:This paper introduces SkillOpt-Lite, a minimal viable pipeline for autonomous agent skill optimization that replaces complex algorithmic architectures with a file-system-based trajectory exploration approach. By treating rollout trajectories as independent flat files and using primitive coding agent tools for consensus mining and validation gating, the framework achieves faster convergence and superior performance across multiple benchmarks compared to heavily engineered baselines.

DSpark: Confidence-Scheduled Speculative Decoding with Semi-Autoregressive Generation

AI Lab:Peking University and DeepSeek-AI

Summary:DSpark is a speculative decoding framework that combines a semi-autoregressive architecture to mitigate acceptance decay with a hardware-aware confidence scheduler to dynamically tailor verification lengths based on system load. When deployed under live user traffic in the DeepSeek-V4 serving system, this approach significantly improves accepted sequence length and accelerates per-user generation speeds without degrading aggregate throughput.

Nemotron-Labs-Diffusion: A Tri-Mode Language Model Unifying Autoregressive, Diffusion, and Self-Speculation Decoding

AI Lab:NVIDIA, Georgia Tech, HKU, University of Chicago, and MIT

Summary:This research presents a tri-mode language model that harmonizes autoregressive and diffusion objectives within a single architecture, allowing dynamic switching between causal, parallel diffusion, and self-speculation decoding modes. The resulting Nemotron-Labs-Diffusion model family consistently outperforms state-of-the-art open-source alternatives by maximizing generation throughput and maintaining high accuracy across various deployment constraints and concurrency levels.

Urban congestion relief experiments through routing-app interventions

AI Lab:Google Research, UC Berkeley, and Stanford GSB

Summary:This paper details a large-scale empirical study across 10 major US cities where a small proportion of Google Maps trips were algorithmically rerouted away from highly congested road segments to less congested alternatives. The intervention yielded an average 2% increase in vehicle speeds on targeted segments and improved overall network travel times, demonstrating that marginal routing adjustments can significantly enhance road efficiency and reduce CO2-equivalent emissions.

烙 AI Tech Releases

GPT 5.6

OpenAIunveiled GPT 5.6which pushes the frontier of AI models.

ChatGPT Work

OpenAIreleased ChatGPT Work, its new agent for productivity workflows powered by GPT 5.6.

Grok 4.5

SpaceXAI’sreleased Grok 4.5with new agentic, coding and knowledge work capabilities.

Muse Spark 1.1

Metaintroduced Muse Spark 1.1, the second version of its multimodal reasoning model.

GPT-Live

OpenAIalso released GPT-Live, its new generation of audio models.

10 AI News You Need to Know About

Lovable is reportedly in talks to raise $300 million at a $13.2 billion valuation, exactly double the mark from its December round, in a deal expected to be led by Menlo Ventures (Sifted broke this one, so it stays the original source since there’s no company announcement yet).
Prime Intellect announced a $130 million Series A led by Radical Ventures, at a reported $1 billion valuation, to build its “open superintelligence stack” of compute, RL post-training, environments, and evals that lets enterprises like Ramp train their own models instead of renting frontier APIs.
SambaNova completed the first close of a $1 billion Series F at an $11 billion post-money valuationled by General Atlantic, alongside news that JPMorganChase will deploy its SN40 and SN50 systems for on-prem inference.
Memory chipmaker CXMT will open investor subscriptions next week for its roughly $4.3 billion Shanghai STAR Market IPO, offering 6.688 billion shares in what is the year’s most anticipated Chinese listing (kept the Bloomberg link since the underlying source is a Chinese exchange filing).
Positron is in talks to raise new funding at a valuation of around $5 billionas the inference-chip challenger to Nvidia seeks roughly $750 million across two phases (Bloomberg scoop on private talks, no press release exists).
Paradigm announced its fourth fund, a $1.2 billion vehiclethat formally expands the crypto-native firm’s mandate into AI, robotics, and other “steep exponentials,” with early bets already in Zipline, True Anomaly, and Nous Research.
Norm Ai raised a $120 million Series C at a $1.2 billion valuation led by Khosla Venturesto scale its “agentic law” model, which pairs AI agents with an affiliated AI-native law firm and supervisory agents for regulated deployments.
OpenAI is sunsetting its Atlas browser less than a year after launch, redistributing its agentic browsing capabilities into a new ChatGPT Chrome extension and a beefed-up desktop app with an embedded browser plus a remote cloud browser for agent tasks — a concession that the browser is a feature, not a destination.
Gradium extended its seed round to $100 million with new investors including NVIDIA, roughly $30 million on top of the $70 million the Kyutai spinout raised at launch in December, and is opening a San Francisco Bay Area office to scale its ultra-low-latency real-time voice AI models.
Ollama announced a $65 million Series B led by Theory Ventures, with Benchmark, 8VC, and Y Combinator participating, bringing total funding to $88 million as the open-model platform hits 8.9 million monthly developers (doubled since January), ~1 million installs a week, and presence in 85% of the Fortune 500 — all with 14 employees.

",{"_256":202,"_257":286,"_258":287},"7f449a18-d88e-4687-95e7-7898ec87927d",{"_21":286,"_33":288,"_204":289,"_35":290,"_226":291,"_263":291,"_23":233,"_264":292,"_266":293,"_219":42,"_70":-5,"_72":42,"_182":294,"_256":202,"_283":307},"Grok x Cursor","Anthropic has extended Fable 5 access for Claude subscribers until July 12, after which prepaid credits will be needed. OpenAI is rolling out GPT-5.6 models (Sol, Terra, Luna) today, with early feedback suggesting Sol is highly reliable. New developments include the Grok 4.5 model, enhanced ChatGPT Voice with new models, and a lighter version of Claude Cowork.","More Fable, GPT-5.6 and new ChatGPT voice","https://substack-post-media.s3.amazonaws.com/public/images/c4128583-dcdf-46c8-8d5a-5b0878ab4434_1800x1200.png","2026-07-09T13:02:45","https://www.bensbites.com/p/grok-x-cursor",{"_21":295,"_185":296,"_187":116,"_70":297,"_213":304,"_214":305,"_216":306},"c83edb190d40b121d8ba13473663a79e5c88cc7aa6003ed628c44fcd69d1eb57","f6e589e625d9d79a6660ea0e37b9d23db53cde5c31ff44b1a2e420da9e9ca3379648885583cdda58704010bc212befa9886ebd0d780b6e50e1598fccfbf4d4a1",[298,299,300,302,303],[219,288],[37,291],[275,301,42,277],"30023:0b16efb3db6165f2311a31e04c8f874f8c569a4cc6cc219a3295cdfa9110b8b8:7f449a18-d88e-4687-95e7-7898ec87927d",[210,42],[212,304,42],"0b16efb3db6165f2311a31e04c8f874f8c569a4cc6cc219a3295cdfa9110b8b8","Grok x Cursor\nMore Fable, GPT-5.6 and new ChatGPT voice\n\n\nnostr:naddr1qvzqqqr4gupzqzcka7eakct97gc35v0qfj8cwnuv26dye3kvyxdr99wdl2g3pw9cqqjrwe35xsukzvfc94jrswr9956rvwph95un2efh95mnswfcv43nsdeexgmkg5dxmm9 ",1783602165,"

Hey folks,

Anthropic extended access to Fable 5 via Claude subscriptions until July 12. Same deal as before: up to 50% of your weekly limits, and after July 12, use prepaid usage credits. But this dangling-a-carrot move doesn’t make me trust their deadlines anymore.

Otoh, OpenAI announced that GPT-5.6 models (Sol, Terra and Luna) will be live for all users later today. General feedback on X (from early testers) is that Sol is not smarter than Fable but is so much more reliable that you would not want to use anything else.

Well, let’s see.

Ben’s Bites is brought to you byFlodesk

Most AI tools generate emails for you.Flodesk Studiodesignswithyou — describe what you want, get a beautiful on-brand email in seconds, then refine in chat or perfect by hand. Real designer foundations, AI-accelerated, finished by you.Try it.

Headlines

We have a good Grok model - Grok 4.5. SpaceXAI and Cursor trained it together, and it’s an Opus-class model. It should perform roughly in between Opus 4.7 and 4.8. But it’s much cheaper on a per-token basis and token-efficient—making it 6x cheaper than Opus models and 3x cheaper than GPT-5.5. It’s available in Cursor with increased limits andvia API too.
ChatGPT Voice is finally getting some new models -GPT-Live-1 and Live-1-mini.A new architecture lets these models a) talk while you’re still speaking without cutting your message off and b) use GPT-5.5 in the background. Results in much smoother conversations and a big jump in intelligence from the old advanced voice mode.
A lighter version ofClaude Cowork will work on web and mobile- sessions will run remotely and save to your account, so you can kick off a task on the laptop, close it, and check in from your phone. Cowork on the desktop app remains the full deal. Beta rolls out over the coming weeks.
Muse Image and Video- New multimedia generation models from Meta. The images have more of an AI look, but the model seems good at image editing and text rendering, and can search the web to get facts right, etc. They also have a video model in preview that will be released soon.

Subscribe now

My feed

Render Workflows- Definelong-runningtasks with a lightweight SDK thathandles queuing, orchestration, & retries.Try it- use codeRENDER-BENSBITESfor $50 in credits.*
Cognition released SWE-1.7- its strongest model yet, near the top frontier models at lower cost, and available at 1000 tok/s.
Cloudflare Drop- drag a folder into the browser, get a live site, no account needed.
Toya- Executive assistant that lives in iMessage, calls your phone, knows you and your work. (i’m an investor; check them out onProduct HuntandTwitter)
Ora Directory- search the web for tools and resources agents can use.
64 Claudesrewrote Bun in Rustin just 11 days.
Spatial contextin Clicky - draw on your screen so the model knows exactly where to look.
LoginWithChatGPT- unofficial SDK to add “Sign in with ChatGPT” in your app.
How to become anapplied AI engineer.
Model vs effortin Claude Code: “big model low thinking” or “small model high thinking”, what should you choose and when?
How to turn raw Markdown notes into anLLM knowledge base.
Acollection of small workershosted on Notion to sync Twitter bookmarks, Readwise, read webpages in markdown and more. (repo)
Perplexity Teammate- reportedly an internal coding agent like Claude Code/Codex.
Skills collections:
- mattpocock/skills v1.1- /wayfinder, /research, /implement, /to-spec and /to-tickets.
- Meng To’s Agent Skills- 75 design, motion, WebGL and landing-page skills for agents.
Afters

annals-kingdom.netlify.app\",\"username\":\"emollick\",\"name\":\"Ethan Mollick\",\"profile_image_url\":\"https://pbs.substack.com/profile_images/1601382188712398850/3AAOlqrX_normal.jpg\",\"date\":\"2026-07-08T01:29:51.000Z\",\"photos\":[{\"img_url\":\"https://substackcdn.com/image/fetch/$s_!_9w3!,w_1028,c_limit,f_auto,q_auto:best,fl_progressive:steep/l_play_button_usfui2,w_88,e_colorize:0/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F__ss-rehost__tw-video-preview-13_2074666913116893184.jpg\",\"link_url\":\"https://t.co/zt7t9nQlzJ\"}],\"quoted_tweet\":{},\"reply_count\":109,\"retweet_count\":205,\"like_count\":4139,\"impression_count\":365500,\"expanded_url\":null,\"video_url\":\"https://video.twimg.com/amplify_video/2074666913116893184/vid/avc1/1174x720/AZmglOuQm8v-YfBj.mp4\",\"belowTheFold\":true}\" data-component-name=\"Twitter2ToDOM\">

Share Ben's Bites

Find me onX,Linkedin, orYouTube
Readabout meand Ben’s Bites
thumbnail via@keshavatearth

* sponsors who make this newsletter possible :)
Wanna partner with us for the next quarter?
Email us atshanice@bensbites.comork@bensbites.com

Ben's Bites is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

",{"_256":202,"_257":309,"_258":310},"0982e74f-173b-41c6-9295-2c0756013f5a",{"_21":309,"_33":311,"_204":312,"_35":313,"_226":314,"_263":314,"_23":233,"_264":315,"_266":316,"_219":42,"_70":-5,"_72":42,"_182":317,"_256":202,"_283":330},"GPT-5.6 is here but...","Etched is launching advanced inference clusters with extreme vertical integration, achieving significant traction with $800M raised and over $1B in backlog orders, aiming to solve the critical bottleneck in AI inference hardware. OpenAI's GPT-5.6 is reportedly ready but faces government blocks, with only select partners gaining access to its new models like Sol. The AI landscape is buzzing with new tools and updates, including Replit's desktop app, in-browser ML models for data redaction, and various agent-building platforms.","plus the insanity of inference","https://substack-post-media.s3.amazonaws.com/public/images/cbc67bbb-9d83-40d2-be1c-c0bab7c9c88e_1200x630.jpeg","2026-06-30T15:02:46","https://www.bensbites.com/p/gpt-56-is-here-but",{"_21":318,"_185":319,"_187":116,"_70":320,"_213":304,"_214":328,"_216":329},"2d5f8616f999d10a76a6ec2ca152c7ceadde39be3e8a0abc98b6f0258d0eef91","cb3822e31f89c1820c26d691db2daad05ed87071d14934c7de6b8863c959f36a99e77d1595bdd66524e2c3faf08af565f1556c1997199b0db2d3a258f9b96b28",[321,322,323,325,327],[219,311],[37,314],[275,324,42,277],"30023:0b16efb3db6165f2311a31e04c8f874f8c569a4cc6cc219a3295cdfa9110b8b8:0982e74f-173b-41c6-9295-2c0756013f5a",[210,326],"Layer3.press",[212,304,42],"GPT-5.6 is here but...\nplus the insanity of inference\n\n\nnostr:naddr1qvzqqqr4gupzqzcka7eakct97gc35v0qfj8cwnuv26dye3kvyxdr99wdl2g3pw9cqqjrqwfcxfjnwdrx95cnwvmz956rzcek95unywf495exxvphx5mrqvfnvc6kz9vnd9c ",1782831766,"

Hey folks, I’m back from a long weekend in Greece and have to tell you about a company I backed a couple years ago that’s coming out of stealth.

Etched.

Training got all the attention for the last few years. But in 2026, the real fight is serving intelligence: lower latency, lower cost, less power, more tokens. The market is screaming for better inference systems, and we’ve never seen a market like inference. So much so, people can’t sleep thinking about it (see the tweet in afters ).

AI inference could be the biggest market in history and the bottleneck is inference hardware. Etched built the perfect product. Pulling it off will change how models run.

Etched are building frontier inference clusters with extreme vertical integration: chips, racks, software, manufacturing, and production co-designed end to end. First product in under three years, while most hardware peers took 7+.

Their traction is already insane: $800M raised, $1B+ in backlog orders, production underway with TSMC, and first chips working on the very first try (A0) on TSMC 4nm. A0 means the first physical version of a chip/system design that comes back from the fab. So yeah, A0 success ishard. Doing it this fast iswild.

I backed Gavin in 2023, as a 21-year old Harvard dropout, because he had first-principles technical conviction and the ability to recruit the most insane team who had actually built the last generation of compute infrastructure. Now Etched is 400+ people from NVIDIA, Google TPU, Broadcom, SK Hynix, TSMC and basically every serious AI chip program.

They’re backed by VentureTech Alliance (which has a strong partnership w/ TSMC), Peter Thiel, Jane Street, Two Sigma, Jump, HRT, Stripes, Ribbit and more.

As we’re seeing, who wins in AI won’t just be decided by who has the best models. But by who can serve them.

Ben’s Bites is brought to you byRender

Chainable compute. Right on queue.
Define tasks with Render’s lightweight SDK and chain them into long-running, distributed workflows. Launch your agents and batch jobs on demand.Render Workflowshandles queuing, orchestration, and retries.
Try it: Use code RENDER-BENSBITES for $50 in credits.

Headlines

OpenAI has GPT-5.6, but its launch is also blocked by the US govt. Only “select partners” are getting access to GPT-5.6 Sol, Terra and Luna, three new models in the family. Sol is the biggest/smartest one of the bunch. It bypasses Mythos in certain benchmarks but is conveniently a little shy of Mythos-level in exploiting cybersecurity bugs. Sam Altman claims that GPT-5.6 will be available to regular users soon—though it might be US only to begin with, even if he’sworking hard for worldwide.
OpenAI also released an economics paper based on theadoption of Codex inside & outside OpenAI. Non-technical adoption is quickly matching engineering departments. The lead of the Codex app was onLenny’s podcast, where he talked about designing the app and the features that matter.
Cursor for iOSlets you launch always-on cloud agents from your phone & remote-control agents running on your computer. Composer 2.5 is 75% off in the app through July 5.
X launched a hosted MCPso Grok, Cursor and other MCP-compatible tools can connect to the X API and X developer docs without building the server yourself.
How do you build AI that actually understands you?Working Smarter, a podcast from Dropbox about AI and modern work, is back for season three. From context engineering to multimodal search, hear how engineers are building AI that works wherever you do.Listen to the first episode now.*

Subscribe now

My feed

Replit has a desktop app nowfor both Mac and Windows.
US national design studio createdRampart- a 14.7MB in-browser ML model to redact PII before it’s sent to a server.
In-app custom agents work for regular users;power users want skillsthey can plug into their own agents to use your product.
Human in the /loop- how do you reliably get in and out of the way of agents?
Zaro- build live apps, agents and workflows from Slack, email, docs and calendars.
Unpeel- native Mac terminal for agents with persistent sessions and git worktrees. (demo)
Tau- educational agent harness for building TUIs, extensions and harnesses.
Inference.netlets you test GLM 5.2 on mirrored prod traffic before switching models for your live users.
Odessia Travel- opinionated AI trip planner that searches and books flights, stays and activities.
smolmachines- spin up hardware-isolated Linux microVMs.
Animation vocabulary- skill for asking AI for motion with words like morph and rubber-band.
Every time you think you need a dashboard, ask your agent to make a throwaway HTML.
Teaching agentsproduct designat Vercel.
MCPs, APIs and CLIs- These are not three separate concepts.
New set of shadcn componentsfor building chat interfaces - streaming chat, scrolling, messages and attachments.

Afters

@ajambrosino) leads the team behind the Codex desktop app at@OpenAI. Codex usage has 6x'd since February, reaching over 5M weekly active users, and nearly 100% of OpenAI's employees use the Codex app regularly (and not just the engineers).\\n\\nAndrew's personal\",\"username\":\"lennysan\",\"name\":\"Lenny Rachitsky\",\"profile_image_url\":\"https://pbs.substack.com/profile_images/1592990461517389824/iln8hi1f_normal.jpg\",\"date\":\"2026-06-28T18:07:04.000Z\",\"photos\":[],\"quoted_tweet\":{},\"reply_count\":36,\"retweet_count\":35,\"like_count\":375,\"impression_count\":363604,\"expanded_url\":null,\"video_url\":null,\"belowTheFold\":true}\" data-component-name=\"Twitter2ToDOM\">

Share Ben's Bites

Find me onX,Linkedin, orYouTube
Readabout meand Ben’s Bites
thumbnail via@keshavatearth

* sponsors who make this newsletter possible :)
Wanna partner with us for the next quarter?
Email us atshanice@bensbites.comork@bensbites.com

Ben's Bites is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

",[],"selectedArticleImage","rootArticleImageUrl","consensusHtml","tweets",[],"userSchema","relatedArticles",[259,287,310],"actionData","errors"]

OpenAI Begins Rollout of GPT-5.6 Model Family

Story-Berichterstattung