Cascadity Audio
← All streams

#techcrunch-disrupt

Everything tagged techcrunch-disrupt, across every stream.

0

Laplace Intelligence red-teams robots and self-driving cars in simulation

Stress-testing physical AI before it ships, from researchers whose method made the cover of Nature.

Ann Arbor startup Laplace Intelligence, founded a year ago by Howie Sun and his former University of Michigan PhD adviser Henry Liu, builds high-fidelity simulations reconstructed from the real world plus an AI agent that hunts for failure points in robots and autonomous vehicles. Their 2023 paper on the approach was a Nature cover story. The company says it already earns revenue, mostly from autonomous driving customers such as trucking firm Bot Auto and a major Japanese carmaker, and is running a smart-city traffic pilot in San Jose with Nvidia and T-Mobile.

Cascadic Analysis 1
UndertowBenchmark scores can overstate real-world reliability

What hidden risk could pull against this, even if the news is good?

Laplace stress-tests robots in simulations reconstructed from the real world. Sun himself says "no one knows what the edge cases will be in the real world"; passing a simulated test suite is evidence, not proof.

0
Rabbit Holes 1
0

Ricursive Intelligence wants AI to design the chips AI runs on

The founders behind AlphaChip explain why faster chip design, not more data centers, is their answer to the compute crunch.

Ricursive Intelligence founders Anna Goldie and Azalia Mirhoseini, who previously developed the AlphaChip system, told TechCrunch Disrupt their startup is building AI that designs chips, learns from each iteration and improves the next design. Goldie said compute demand could grow at least 10,000-fold in five years and argued for more efficient chips over more data centers. They say they optimize whole chips rather than splitting them into dozens of blocks, and that a one-week design cycle would make custom chips for niches like AR, space and biological computing worthwhile. Results beyond AlphaChip have not yet been published.

Cascadic Analysis 2
UndertowWe don't completely understand why frontier models behave as they do

What hidden risk could pull against this, even if the news is good?

The founders say AI finds layouts engineers wouldn't, like AlphaChip's "donut-shaped" arrangement that surprised people before it worked. When a design works for reasons nobody fully understands, it is also hard to know when and why it might fail.

0
UndertowDelayed Consequence / False Success Problem

What hidden risk could pull against this, even if the news is good?

A one-week design cycle means more custom chips reach manufacturing with less human time spent on each. Chip flaws are expensive to find late; problems that slip past automated checks may only appear after the silicon is made.

0
Rabbit Holes 1
0

Apate sends AI agents undercover to get scammed

A defensive use of AI agents: wasting scammers' time while collecting the bank accounts and wallets they use.

Australian startup Apate.ai, spun out of Macquarie University research, runs thousands of AI agents that pose as potential victims: they answer scam calls, reply to phishing messages and join scam groups, in many languages and voices. The company says its agents have wasted 17 years of scammers' time across 2.5 million conversations, and that the goal is to extract bank accounts, crypto wallets, phone numbers and impersonated brands, which it sells to banks, telcos, governments and utilities. Apate was a Startup Battlefield finalist at TechCrunch Disrupt; the article cites FBI figures of more than 22,000 AI-related crime complaints in 2025 with about $893 million in losses.

Cascadic Analysis 2
UndertowRole underspecification / the Doorman Problem

What hidden risk could pull against this, even if the news is good?

Apate's agents pretend to be victims and follow phishing links on purpose. An agent built to look gullible has to know exactly where to stop: which links to open, what to type, and what never to hand over, even while acting the part.

0
UndertowAutonomous cyber capability scales attackers enormously

What hidden risk could pull against this, even if the news is good?

The article cites more than $632 million in 2025 US losses to investment scams where the FBI says fraudsters use AI to run conversations. Apate's thousands of agents show how cheaply conversation can be scaled; the scammers have the same tools.

0
Rabbit Holes 1
0

Industrial Next wants factory robots that switch tasks in minutes

Two former Tesla factory engineers explain why retooling, not robot arms, is the slow part of automation.

San Francisco startup Industrial Next, founded in 2021 by former Tesla engineers Lukas Pankau and Allen Pan, builds self-contained robotic stations driven by its own foundation models linking vision, task context and motion. It says a production line can be retooled in minutes instead of weeks without traditional programming. On the TechCrunch Disrupt Startup Battlefield stage the company announced it is working with BMW Group's Technology Office to evaluate its dexterous manipulation for manufacturing.

Cascadic Analysis 1
UndertowLoss of meaningful human control as capability and autonomy increase

What hidden risk could pull against this, even if the news is good?

Retooling a line in minutes "without traditional programming" means fewer engineers who can read exactly what a station will do. On a factory floor, a model's wrong guess moves metal, not text.

0
Rabbit Holes 1
0

Helix Carbon wants to clean up blast furnaces instead of replacing them

A decarbonization pitch built around the equipment steelmakers already own.

Helix Carbon's electrolyzer turns the CO2 in a blast furnace's raw exhaust back into carbon monoxide, which is fed back into the furnace to reduce iron ore, cutting how much coke the furnace burns. CEO David Brown says the design can handle the full exhaust stream without purification. The company proposes starting by replacing 33% of coke, which it says saves $14 per ton of iron, and plans to sell to specialty chemical makers first to bring in revenue sooner. It pitched in Startup Battlefield at TechCrunch Disrupt.

Rabbit Holes 1
0

Blueprints AI says it can generate permit-ready building plans in minutes

An architect-turned-founder claims AI can take over the drafting that fills most of an architect's day.

Santa Monica startup Blueprints AI, founded by architect and contractor Al Dram with Coco Cao, says its software turns a 2D sketch or PDF into code-compliant, permit-ready construction drawings in minutes. Dram says architects spend about 80% of their time drafting and 20% designing, and wants to flip that ratio. The company, part of TechCrunch Disrupt's Startup Battlefield 200, planned to unveil a second-generation version at the event. Claims about permit acceptance come from the company.

Cascadic Analysis 2
UndertowDelayed Consequence / False Success Problem

What hidden risk could pull against this, even if the news is good?

"Permit-ready" drawings are judged by permit reviewers, not by the people who live in the building. A code violation or structural mistake that passes review may surface only years later, in a finished house.

0
UndertowSpecification gaming / goal misgeneralization

What hidden risk could pull against this, even if the news is good?

The software is built to produce plans that are code-compliant and approvable. Passing a permit check is a proxy for a good, safe building; a system optimized for the proxy can satisfy it without serving the goal.

0
Rabbit Holes 1
0

CedarDB launches a cloud version of its transactions-plus-analytics database

An academic database spinout argues AI agents asking questions in plain English make the old split between database types costly.

Munich-based CedarDB, a Technical University of Munich spinout built on the Umbra research engine, runs both transactional and analytical workloads in one database and is compatible with PostgreSQL, which CEO Moritz Sichert says keeps switching costs low and suits how LLMs write queries. After launching an on-premise enterprise product and a free community edition in 2025 alongside a €5.3 million seed round, the company announced limited availability of CedarDB Cloud at TechCrunch Disrupt.

Cascadic Analysis 1
UndertowConfused-deputy / excessive-agency problem

What hidden risk could pull against this, even if the news is good?

CedarDB's pitch is that AI agents can ask analytical questions of the same database that holds live transactions. An agent that can read the books is close to one that can write them; how narrowly its access is scoped matters more than ever.

0
Rabbit Holes 1