Facts you can check on October 8
- TechCrunch Disrupt 2026 is October 13–15 at Moscone West in San Francisco. The October 5 Disrupt Stage piece lists 10,000+ attendees, 200+ sessions, 250+ speakers, and 300+ exhibiting startups. Event page: the event page. Stage lineup: the October 5 lineup.
- On the developer-tool side the main stage has already named Replit, Cerebras, and Ricursive Intelligence. The October 6 breakout story says two Anthropic engineers will show Claude agents running in parallel. Breakouts: the October 6 agenda.
- Keep the funding figures apart. The September 24 preview says Shield AI raised a $1.5 billion Series G at a $12.7 billion post-money valuation, and Waabi raised $1 billion in January. The Battlefield winner gets $100,000 with no equity. EasyMate’s company page, as of August 31, 2026, lists $4.7 million.
- Startup Battlefield 200 was published on August 20. Company pages you can open, and that sit next to agents or a developer workflow, include Dosu, Collar AI, GuideAI, and Intelligible. The top 20 are not named until October 13. List: Startup Battlefield 200.
- As of October 8, 2026, TechCrunch has not posted a same-day product catalog. The October 5 piece only says Amazon’s Panos Panay will bring some new tech. It does not name the product.
Read developer tools and startups as separate layers. Replit, Cerebras, and Anthropic are already the reference points on stage. Dosu is still raising. Putting Shield AI’s $12.7 billion post-money valuation on the same list as a seed-stage booth company mixes stages.
Five days out: the clock, and the names that are still blank
The dates are October 13–15 at Moscone West in San Francisco. Mid-October is still Pacific Daylight Time (UTC−7). Beijing is 15 hours ahead. The October 13, 10:30 a.m. session on competing for AI talent is 01:30 on October 14 in Beijing. The October 14, 9:55 a.m. edge-AI session is 00:55 on October 15. The October 15, 1:00 p.m. session on reaching $10 million ARR in 90 days is 04:00 on October 16.
Tickets are still being sold in the October 6 breakout story: up to $100 off before doors open, a second pass at half price for a co-founder or colleague, and a further group rate for four or more. People recently laid off can buy a $75 Expo+ pass. The last window for an exhibit table was October 2, which has already passed. Sessions sit across six industry stages, roundtables, and breakouts. The August 20 Battlefield story lists Google for Startups and SAP as premier sponsors, and puts alumni fundraising above $32 billion with 250 exits.
As of October 8, neither the October 5 stage story nor the October 6 breakout story promises a free full-conference livestream. If you are not in San Francisco, the names and sessions already on the agenda are what you can lock in. The top 20 wait until the 13th. Booth demos will mostly show up in post-event copy. Do not assume one official feed.
Developer tools: companies the stage has already named
The table keeps only developer-tool sessions whose names and roles appear in the October 5 stage story, the October 6 breakouts, the agenda, or the September previews. Funding and user figures are TechCrunch’s sentences. Stage lineup: October 5. Deploying Claude: the September 28 preview. Agenda: the Disrupt agenda.
| Company | What it has to do with developer tools | What the agenda already says |
|---|---|---|
| Replit | Turning an idea into software that runs | The October 5 piece: Amjad Masad on what happens to software, and to the people who build it, when almost anyone can create software. The August 24 preview also says the conversation covers finishing a product after the prototype, not only revenue. Preview: Masad joins the Disrupt Stage. |
| Cerebras | Inference infrastructure, not an editor | Andrew Feldman asks whether AI can keep scaling: compute, energy, and capital. In the October 6 breakout, Hagay Lupesko and Rudina Seseri of Glasswing talk about the inference economy: the model itself stops being the moat, and latency becomes part of the product. |
| Ricursive Intelligence | AI that designs the chips it runs on | Anna Goldie and Azalia Mirhoseini. The October 5 piece says the old split between software and silicon may collapse. That is not an IDE. |
| Anthropic | How Claude enters an enterprise workflow, and how the company hands engineering to agents | September 28: Cat de Jong, head of applied AI, on where deployments succeed, where they stall, and why some organizations are still in pilots 18 months later. October 6: Steve Androulakis and Sachin Malhotra on running several Claude agents in parallel, delegating real engineering work, reviewing output, and recovering when an agent is wrong. |
| Clay | Data access, agent workflows, and go-to-market | The September 28 preview names co-founder and CEO Kareem Amin. The same piece says TechCrunch reported in January that Clay was one of the launch apps inside Claude’s workplace tools. |
| The Google Cloud session | Getting an agent from a prototype to a paying customer | The agenda title is The Builder’s Playbook. Stephan Bitton covers unit economics, architectural flexibility, and using cloud distribution to put agents in front of paying enterprise customers. It is a method session, not a startup launch. |
Replit and Anthropic are about how developers hand software off each day. Cerebras is about the inference bill and latency. Ricursive is about chips; do not file it under editors. The speaker list also includes NVIDIA’s Nader Khalil, Director of Developer Tech. That list does not give his session a title, so this piece does not treat it as a developer-tool launch.
AI agents: the workflow on the breakout agenda, and booth companies you can check
The October 6 breakout story frames this year’s questions in five lines: how you build with AI agents; whether value moves from training to inference; what physical AI actually is; how you run a fundraise; and how AI changes growth and teams. The first two are the ones that sit next to developers. The Anthropic session, in that story’s words, is what an agent workflow looks like when the people using it build AI for a living. They will show parallel runs, delegation, review, and recovery. That is a workflow, not a product list.
The early booth companies come from the August 20 list of 200. The enterprise-tech slice includes Dosu, Collar AI, GuideAI, Intelligible, EasyMate AI, Latch AI, and Spiky.ai, among others. The table keeps only companies whose TechCrunch page opened with a sentence you can check. Kilo is on the list, but there is no checkable company-page sentence, so this piece does not guess what it does. Dosu: TechCrunch company page.
| Company | The agent-related sentence on the company page | Status on that page |
|---|---|---|
| Dosu | Knowledge infrastructure for agents. Founder Devin Stein, California. The company site says it connects GitHub or GitLab and Slack, and offers an MCP server so a coding agent reads docs that stay current. Site: dosu.dev. | Open, currently raising |
| Collar AI | A multi-agent system for private-market investors: finding deals, winning them, and keeping the fund’s information. Founders Jamie Yau and Kazel Koh. | Currently raising |
| GuideAI | Voice-driven client guidance for regulated industries. Founder Phil Hong. | Currently raising |
| Intelligible | A context engine for AI systems that work on tabular data. Founders Ben Lengerich, Rich Caruana, and Joshua Shunk. | Open; the page does not mark it as currently raising |
| Latch AI | Catches customers who are quietly leaving. Founder Alisha Esmail, Canada. | Currently raising |
| Spiky.ai | The company page uses the line “the world’s first EQ foundational model.” That is their positioning, not a third-party evaluation. Founders Furkan Eris and Burak Aksar, California. | Currently raising |
| EasyMate AI | Customer-facing personalized engagement. Founders Aihui Ong and Michael Han. | As of 2026-08-31 the page lists $4.7 million raised |
These seven sell to different buyers: a knowledge layer for coding agents, a multi-agent workflow for private markets, voice guidance in regulated industries, context for tables, retention, an emotion model aimed at sales, and customer engagement. The agenda also lists Anthropic’s Kevin Bai on forward deployed engineering. That is an enterprise sales motion, not another booth company. Before October 13 nobody knows who reaches the top 20. The useful check is whether the workflow matches yours.
New products: what was previewed, and what still has no name
Searching for “what launches at Disrupt 2026” still does not land on a catalog. As of October 8, the event page, the October 5 stage story, and the October 6 breakout story do not list products shipping that week. What they do state is live demos, Battlefield pitches, and a $100,000 equity-free prize plus the Disrupt Cup.
Read “new product” in three buckets:
- Already happened, and likely to be used as a case. The September 28 preview cites a March TechCrunch report that Gamma was approaching 100 million users, and that the product had moved from presentations into marketing visuals. The same piece says Clay is inside Claude’s workplace tools. The September 24 preview covers Shield AI’s Hivemind selection and Series G, and Waabi’s $1 billion round and Uber partnership. None of these first ship on October 13. Physical AI session: building AI when failure is not an option.
- Named, but not a developer-tool launch. The October 5 piece says Amazon’s Panos Panay will ask what comes after the smartphone, and that he will bring some new tech. The product name is not in that story. The same piece says Nabiha Saklayen of Cellino, who won Battlefield in 2021, will unveil a new product and ask whether you can bank your biology. That is biotech, not an IDE or an agent framework.
- A first demo on the floor whose name has not been previewed. The August 20 Battlefield story says “live demos of never-before-seen tech.” Those names wait for the floor on October 13. They cannot be filled in on October 8.
The prototype-to-production session names Boris Sofman of Bedrock Robotics (formerly Waymo) and John Mackey of MBRYONICS. Preview: From Prototype to Production. It is about turning a prototype into production, not a SKU list.
Funding: the numbers already written down
The only rounds that belong in the piece are sentences on a TechCrunch preview or a company page. September 24: Hivemind was selected in February as an autonomy provider for the U.S. Air Force’s Collaborative Combat Aircraft prototype, and a month later Shield AI announced a $1.5 billion Series G at a $12.7 billion post-money valuation. The same preview: in January Waabi raised $1 billion and partnered with Uber to deploy 25,000 or more robotaxis on Waabi Driver. Waabi World is its simulator. The company says autonomous trucks still need full validation before the safety driver comes out. EasyMate’s $4.7 million appears only on its company page, dated August 31, 2026. The Battlefield prize is $100,000 with no equity, confirmed again in the October 5 story.
| Judge | Firm | Scope written on the page |
|---|---|---|
| Caleb Appleton | Partner, Bison Ventures | Physical AI: techbio, robotics, and real-world intelligence. Examples include Cobot, Vivodyne, and Grid Aero. |
| Sara Deshpande | General Partner, Maven Ventures | Seed-stage consumer software, digital health, and consumer AI. |
| Aatish Nayak | Partner, Kleiner Perkins | Joined in May 2026 to back AI-native founders with a vertical specialty. Previously the first product manager and later VP of product at Harvey, and before that at Scale AI and Shield AI. |
| Jason Risch | Partner, Greylock | Enterprise security, AI and ML infrastructure, data platforms, and developer tools. The portfolio includes Onehouse and Baseten. |
| Mark Xu | Partner, Index Ventures | Cybersecurity, infrastructure, and AI. Has backed Fireworks, Parallel, 7AI, Simile, and Flapping Airplanes. Previously at Lightspeed, including Wiz, Glean, and Grafana. |
On September 28 TechCrunch named these five and said 20 judges are now public. The other five, who decide the $100,000, are not named in the October 5 or October 6 stories. Source: the next judges. The pages that sit next to developer tools are Risch and Xu. Risch’s page lists developer tools and ML infrastructure. Xu’s lists inference companies such as Fireworks. Appleton’s page is physical AI. That is the scope written on their pages, not a market forecast.
Nayak’s résumé joins “leave the pilot” to “pick a vertical”: Scale’s data, Shield’s defense work, Harvey’s legal product, then Kleiner looking for AI-native founders with a vertical specialty. The $100,000 prize does not reset the funding market. It changes who gets written up on October 14.
Trends: four lines the agenda keeps repeating
These four lines are questions the October 5 and October 6 agenda stories keep returning to. They are not a prediction of who wins. Still unpublished as of October 8: the top 20 on the 13th, the last five final-round judges, and the names of the booth demos.
- Agents move from a demo into daily engineering. The evidence is Anthropic’s breakout, Google Cloud’s Builder’s Playbook, Replit on the main stage, and booth projects such as Dosu that build a knowledge layer for coding agents.
- Value moves from training to inference. Feldman asks whether AI can keep scaling. Lupesko’s session says the model is no longer the moat, latency becomes part of the product, and investors will look next at infrastructure and unit economics.
- Physical AI is execution and verification, not a larger model. Shield AI, Waabi, and the October 14 edge session with FieldAI and Medra are about places the cloud cannot reach. On October 6, Ethan Batraski of Venrock, Ivan Poupyrev of Archetype AI, and Ian Rountree of Cantos have a session whose title is the execution layer: perception, decision, action, verification.
- The checks split between developer tools and verticals. Greylock lists developer tools, Index lists inference infrastructure, Kleiner is looking for founders with a vertical specialty, and Bison backs physical AI. It is not one bet on a larger general model.
Questions
When is TechCrunch Disrupt 2026?
October 13–15 at Moscone West in San Francisco. In Beijing that runs from the afternoon of the 13th into the early morning of the 16th. 10:30 a.m. PDT on the 13th is 01:30 on October 14 in Beijing.
Is there a free livestream?
As of October 8, 2026, the October 5 stage story and the October 6 breakout story are selling tickets. Neither promises a free full-conference livestream. What is confirmed is 10,000+ people on site.
Which AI startups and developer tools are worth watching?
Split them. Already named for the main stage: Replit, Cerebras, and Ricursive Intelligence. For agent workflows, watch Anthropic’s breakout and Clay. Booth companies with a checkable page include Dosu, Collar AI, GuideAI, and Intelligible. The top 20 are not named until October 13.
How are AI agents discussed at the event?
The October 6 breakout says Anthropic’s Steve Androulakis and Sachin Malhotra will show several Claude agents running in parallel, real engineering work delegated to them, output reviewed, and recovery when an agent is wrong. The agenda also lists Google Cloud’s Stephan Bitton on getting an agent from a prototype to a paying customer. Those are workflow sessions, not a product launch.
What new products will launch?
As of October 8 there is no official product list. The October 5 story only says Amazon’s Panos Panay will bring some new tech, with no product name. Cellino’s unveil is about banking biology, not a developer tool. Gamma’s user scale, Clay inside Claude, and the Shield AI and Waabi rounds already happened before the event.
Which funding numbers are already public?
The September 24 preview says Shield AI raised a $1.5 billion Series G at a $12.7 billion post-money valuation, and Waabi raised $1 billion in January. The Battlefield winner receives $100,000 with no equity. EasyMate’s company page, as of August 31, 2026, lists $4.7 million. The five judges named on September 28 cover physical AI, consumer AI, AI-native founders with a vertical, plus enterprise security, ML infrastructure, and developer tools.
What to actually watch before the doors open
Disrupt 2026 is not one blended hot-startup list. For developer tools, watch how Replit, Cerebras, and Anthropic talk about agents. For early companies, watch names such as Dosu that are still raising and whose company page matches the claim. For funding, use only the numbers already written on a preview or a company page. Product names wait for the floor on October 13. Do not fill them in on October 8.
If the question is how models and developer tools are priced and wired into a workflow, read the product news that has already shipped, rather than waiting for a startup conference to announce it: OpenAI DevDay 2026: Which New Products to Watch, and What Is DeepSeek Harness: How dsh Combines Models and Tools.
Pin the agent demo to a Mac that does not close its lid
Dosu’s booth project (a knowledge layer on MCP) and the stage sessions from Replit and Anthropic on parallel agents both assume the machine stays on. The usual pre-show failure is a laptop lid, a drifting environment, and dependencies installed the morning of the demo. kvmboot’s cloud Mac mini keeps that environment in a data center: macOS Unix, Homebrew, and SSH are there on day one, and a long job does not stop because a lid closed. Apple Silicon unified memory lets you run a local model draft before you decide whether the heavy compute belongs on a cloud GPU. An M4 Mini idles at about 4W, which fits a rehearsal box during the show. Stability, the toolchain, and energy use are three separate facts.
Compare a cloud Mac with buying one · See plans · Back to the homepage