DEV Community: Nolan

Can a Cap Pass Your Intentions to AI? What Sabi's $50 Million Seed Round Has to Prove

Nolan — Tue, 20 Oct 2026 12:25:35 +0000

Sabi wants to turn deliberate thoughts into AI commands through an everyday cap. The funding is large and the product is still in development. Here is what the money should produce, and what the evidence shows so far.

The funding

Source: Sabi company website (sabi.com)

On October 9, 2026, Sabi, a non-invasive brain-computer interface startup, announced a $50 million seed round led by Khosla Ventures. Accel, Initialized Capital, and DST Global also took part, among others. Former OpenAI chief product officer Kevin Weil is on the investor list. Forbes reports the company is based in Palo Alto.

The idea is simple to state. Today, when you want AI to do something, you type or speak. Sabi wants a cap that picks up brain signals when you deliberately form an instruction and turns them into input a computer can receive.

Nobody can buy the product yet. That makes $50 million at the seed stage a large check for an unproven device.

What the money should turn into evidence

Where the $50M is going, and the open questions it has to answer

The question for a seed-stage company is which uncertainties the money can resolve. For Sabi, the main ones are:

Can the chip capture signals consistently?
Can the model adapt to different people?
Can the cap be comfortable enough to wear?
Will people want to keep using it?

The announcement says the funds will go toward a larger neural dataset, finishing a custom chip, building early devices, and growing the team. These are the steps between a research prototype and a usable product.

The possible use cases are easy to picture. Your hands might be busy. You might be somewhere you can't talk out loud. But these are potential uses, not proven ones. Only real users can show whether a cap beats typing or speaking.

The team: research is not the product

MindEye and the Sabi cap differ in data, equipment and task

Sabi has two cofounders: CEO Rahul Chhabra and CTO Atmadeep Banerjee. Banerjee is a co-first author of MindEye, a NeurIPS 2023 paper.

That background matters, but it needs context. MindEye used fMRI (functional magnetic resonance imaging) data to retrieve and reconstruct images. It shows the team can combine brain data with machine learning.

Sabi's cap uses EEG (electroencephalography) and is meant for everyday wear. The data, the equipment, and the task are all different. The research says something about the team's ability. It says nothing about how the cap performs.

The same applies to the investor list. It shows Sabi has backing. It doesn't show how much each firm invested, and it doesn't mean any investor has independently validated the product.

How the technology is supposed to work

Sabi's stated three-stage design and the open question at each stage

The system can be broken into three stages.

1. Capture the signal. EEG records electrical signals linked to brain activity. According to the company, Sabi puts custom chips and tiny sensors inside a cap that reads signals through hair. "Non-invasive" here means no surgery, no conductive gel, and no direct scalp contact. This is the company's stated design.

Everyday wear is the hard part. Blinking, facial muscles, and body movement can all distort EEG recordings. Researchers call this interference "artifacts," and handling them is a known signal-processing problem. That is a general challenge for EEG, not a reported flaw in Sabi's system.

2. Decode the intention. Sabi feeds the signals into what it calls a Brain Foundation Model. Its website describes the sequence: sensors collect data, the data trains a model, and the model maps signals to outputs.

The practical questions are specific:

How long does a new user need to calibrate?
Does a model trained today still work when the same person puts the cap on tomorrow?
What happens when fit, hairstyle, or activity changes?

These answers decide whether the device works outside a lab.

3. Hand the result to AI. Once the model thinks it has recognized an intention, when should the system act? A sensible design would show a candidate command and let the user confirm or fix it first. A language model can make a command read smoothly, but smooth output doesn't mean it kept what the user meant. The input layer needs a way to correct and cancel.

The 77 percent figure

The 77% figure and the conditions that limit how it can be read

This is the number most likely to be quoted, so it needs care.

In the Forbes report, the CEO said an early, wired, helmet-like prototype reached up to 77 percent transcription accuracy in lab conditions, for users who had used it for more than two weeks.

It is a specific figure, not a vague claim. But it comes with conditions:

Lab setting
Early, wired, helmet-style prototype, not the cap
Users with more than two weeks of practice
An "up to" result, not an average
A founder's statement to a journalist, without a published test protocol

It cannot be read as the everyday accuracy of a wireless cap. On its own, it also doesn't tell us whether the product is useful. We would need to know how accuracy was measured, whether vocabulary was limited, how many people were tested, and how failures were recorded.

What would make it a business

Counting the whole interaction, and the reinforcing loop that would still need numbers

A first product needs one task people want to repeat. "Everyone needs a more natural way to talk to AI" is a vision, not a use case. The narrower question is where the cap saves time, cuts errors, or makes something possible that wasn't before.

That means counting the whole interaction: putting the cap on, calibrating, waiting, confirming, and fixing mistakes. Fast recognition won't help much if setup takes minutes every time.

Public materials don't show pricing, paying customers, or unit economics. Hardware sales, software subscriptions, or developer access are all possible paths. For now, they are hypotheses.

On defensibility, the question is whether the chip, the data, and the user experience reinforce each other. Better capture could yield better training data. A better model could shorten calibration. A better experience could keep people wearing it. Each of these is a "could" that needs numbers behind it.

A large dataset is not automatically a moat. It matters whether the data covers different users and conditions, whether it reduces errors, and whether data permissions allow long-term training and product use.

The tests that matter

Six repeatable tests a credible evaluation would include

The gap between a good demo and a usable product is closed by repeated testing, not a single showing. A credible evaluation would:

Time a new user's first setup
Bring users back the next day to see if they must recalibrate
Let users move normally and record how long the device works and when it fails
Measure correct commands per minute and the time needed to fix one error
Track false activations, when the system acts without the user meaning to give a command
Have an independent team reproduce results under clearly described conditions

One smooth minute in a demo says little about the rest of the day.

What to watch next

Three questions for Sabi's planned CES 2027 prototype showing

Sabi says it plans to show early interaction prototypes at CES 2027. When that demo happens, three questions matter more than how impressive it looks:

How quickly can a new user get started?
How many errors occur in ordinary conditions?
After correcting those errors, is it still easier than typing or speaking?

Sabi has a direction worth following and the money to pursue it. Whether it becomes a product people keep wearing depends on evidence that doesn't exist publicly yet.

Trmeric Bets an AI Chief of Staff Can Connect Strategy to Execution

Nolan — Sun, 18 Oct 2026 12:56:34 +0000

The San Ramon startup raised a $6.5 million Seed round led by Hitachi Ventures. Its pitch is that AI agents can do the planning, tracking, and ROI work that large companies now spread across slide decks, spreadsheets, and consultants.

The gap between strategy and execution

The strategy–execution split: two sides of the company working in separate tools.

Large companies spend heavily on digital transformation. One estimate puts Fortune 500 spending on transformation, cloud migration, and AI agents at more than $1.5 trillion a year in 2026.

Many of these projects still go wrong. McKinsey and Gartner surveys suggest that nearly 70% of enterprise transformation projects go over budget, run badly late, or fail outright.

A common cause is a split inside the company:

Executives set strategy in slide decks and spreadsheets.
Engineering and IT teams do the work in Jira, GitHub, and SAP.

The two sides rarely share a live view. Leadership often can't see where a project stands or measure what its spending is returning. When something breaks, the usual fix is to hire a consulting firm such as McKinsey or Accenture and pay for many hours of work.

Who Trmeric is and who backed it

Trmeric's $6.5 million Seed round at a glance.

Trmeric is based in San Ramon, California. It wants to close this gap with agentic AI, meaning AI agents that can carry out work rather than just store it.

The company has announced a $6.5 million Seed round:

Lead investor: Hitachi Ventures, the venture arm of Hitachi.
Other investors: Schema Ventures, CerraCap Impact VC, FalconX Ventures, and Executive Venture Fund.
Angel investors: Gokul Rajaram, a former Google and DoorDash executive, plus Yatish Mishra and Shailesh Lakhani.

Trmeric says its platform has already supported more than $1 billion in enterprise transformation projects. Much of that came while the company was still in stealth.

The co-founders are CEO Siddharth Bohra and Debottam Datta. They spent about 20 years as senior executives at IT consulting firms, running large transformation programs at Fortune 500 companies. They saw good strategies stall in status meetings, patched-together decks, and fights between departments. That experience with how consultants charge, and why existing software is rigid, shaped the product.

The investment case: from recording work to driving it

The shift investors describe: from storing work to acting on it.

Investors frame Trmeric as part of a shift in enterprise software:

A System of Record stores what people enter. Most project tools work this way.
A System of Agency acts on that information. It plans, assigns, monitors, and warns.

For Hitachi, the fit may be strategic as well as financial. Hitachi runs large industrial and IT operations through its Lumada platform. Leading the round could give it a way to build Trmeric into the solutions it sells to enterprise clients. How deep that integration will go has not been detailed.

The pricing argument is about replacing consulting spend. A major architecture overhaul can cost millions in consulting fees. Trmeric sells a software subscription instead. According to the company's own data, it:

Frees up as many as 52 productive days per user each year.
Shrinks planning cycles from months to days.

These figures come from Trmeric and have not been independently verified. If they hold up, software that cuts outside consulting costs and can show project ROI should have pricing power and keep customers.

How it works: Tango and its agents

Source: Trmeric website (trmeric.com)

Tools like Jira, ServiceNow, and Clarity are, at heart, static forms. They wait for someone to enter data. They don't turn an abstract goal into concrete tasks.

Trmeric's design has two layers:

A multi-agent layer led by an agent called Tango.
A knowledge layer the company calls the Enterprise Digital Brain.

Tango acts as an AI chief of staff for the executive team. It breaks down strategy, matches people to work, watches for risk, and checks ROI.

Take an illustrative goal: a CIO wants to move supply chain systems to SAP S/4HANA by Q3 and cut latency by 15%. Tango doesn't just split that into a task list. It hands the work to specialized agents:

Planning agent: asks for missing context, then estimates budget, resources, and milestones.
Resource agent: checks skills and open capacity across departments to find the right people.
Execution and governance agent: monitors Jira, GitHub, SAP, and email around the clock for signs of risk.

The initiative priority score

How the four inputs push the initiative priority score up or down.

Tango also scores each project so money goes to work that pays back. This is a form of AI-driven project portfolio management.

The score uses four inputs:

Business alignment
Projected ROI
Capital expenditure (CapEx)
Execution risk

In plain terms, the score is weighted alignment plus weighted ROI, divided by weighted CapEx plus weighted risk:

Priority = (w₁ × Alignment + w₂ × ROI) ÷ (w₃ × CapEx + w₄ × Risk)

Higher alignment and return push the score up. Higher spending and risk pull it down. The company says the score is computed in milliseconds. How the weights are set, and how ROI and risk are estimated, has not been disclosed.

The goal is to flag low-return projects before money is committed, so budget can move to better bets.

The Enterprise Digital Brain

Source: Trmeric website (trmeric.com)

When a project ends, its lessons usually stay with the people who worked on it. Post-mortems sit inside one team.

Trmeric says it automatically captures each project's decision history, risk points, and code architecture in a shared knowledge base. When a new program starts, the agents draw on that history to recommend a path, so teams don't start from zero.

What the company plans next

Where the new funding goes: team hubs and target industries.

With the new funding, Trmeric is expanding its AI engineering and enterprise sales teams in San Francisco, Dallas, and Bengaluru. It plans to move beyond technology and semiconductor customers into large companies in finance, manufacturing, and energy.

The hard part: legacy systems and security

What agents need to act on legacy stacks, and the three linked problems it raises.

The biggest open question is engineering, not funding.

Enterprise IT stacks are full of old SAP and Oracle installations and custom internal systems, some more than a decade old. For Tango to act, its agents need reliable read and write access through APIs across all of them. That raises three linked problems:

Integration: connecting to many different legacy systems.
Security: isolating data and meeting standards such as SOC 2 and CCPA, ideally under a zero-trust model.
Reliability: making sure agents don't make reasoning errors, or "hallucinate," when they reassign people and resources.

How well Trmeric solves these will decide whether it can become core infrastructure for enterprise transformation, or stays a planning layer on top.

The question for transformation leaders

Source: Trmeric website (trmeric.com)

Trmeric pairs founders who know where transformation breaks with backing from Hitachi and Silicon Valley product leaders. Its bet is that enterprise software can move from recording work to driving it.

Much of the evidence so far comes from the company itself, and the hardest integration work is still ahead. But the problem it targets is real. If you run transformation projects or are evaluating AI agents for the enterprise, start with one question: what connects your strategy to your execution today?

Aventra Defense Raises $17.5M to Send Balloon-Launched Glide Payloads Into the Stratosphere

Nolan — Thu, 08 Oct 2026 18:26:36 +0000

The Virginia startup's Piranha platform rides cheap balloons to 60,000–100,000 feet, then glides silently toward targets hundreds of kilometers away. Backers say it addresses the cost problem at the heart of modern air warfare.

The cost problem driving the deal

The cost asymmetry described in the article, using the reported figures. Bars are illustrative, not to scale.

Modern air defense has an arithmetic problem. In recent conflicts, including the war in Ukraine and anti-missile engagements in the Red Sea, defenders have fired interceptors reportedly costing $2 million to $4 million each at attack drones costing around $20,000. That exchange rate drains stockpiles fast, even for well-funded militaries.

The offensive side has a mirror-image problem. Cruise missiles and stealth drones used for long-range reconnaissance and precision strikes run roughly $1 million to $5 million per unit. Satellites can cover the same ground, but they are expensive, vulnerable to jamming and anti-satellite weapons, and locked to orbits that keep them from loitering over one area.

Aventra Defense Systems, based in Herndon, Virginia, is pitching a cheaper path: put sensors, jammers, or warheads in the stratosphere, above the reach of most air defenses, using balloons instead of rockets or jet engines.

The funding round

Participants in the $17.5M round as announced. Original diagram.

Aventra announced $17.5 million in new financing, including a credit facility from Western Alliance Bank. Konvoy led the round. Returning investor Lavrock Ventures participated, along with Booz Allen Ventures, Sands Capital, and Langford Industries.

The Booz Allen connection matters beyond the money. Booz Allen Hamilton is a major systems integrator for U.S. military intelligence, electronic warfare, and C4ISR programs. For a defense startup, the hardest stretch is the "valley of death" — working technology but no procurement contract. An investor with integration pathways into Army and Marine Corps multi-domain programs could shorten that gap, though an investment is not a contract.

The round also fits a broader Pentagon push, often associated with the Replicator initiative, toward low-cost "attritable" systems — weapons cheap enough to lose in quantity.

How Piranha works

Piranha's two-stage mission profile as described by Aventra. Original diagram, not to scale.

Aventra was founded in 2025 by Michael Weigand, Brian Retherford, and Jessup Meng, a team with backgrounds in military drones, autonomous navigation, and high-altitude platforms. Their Piranha platform splits the mission into two stages.

Stage one: balloon lift. A helium- or hydrogen-filled balloon, reportedly costing a few thousand dollars, carries the Piranha payload to between 60,000 and 100,000 feet. At the planned release point, a thermal cutoff detaches the vehicle from the balloon.

Stage two: unpowered glide. Piranha unfolds high-aspect-ratio wings and glides. Air density at that altitude is roughly 1% to 7% of sea level — too thin for conventional aircraft to generate useful lift, but workable for a purpose-built glider. Range depends on release altitude and the vehicle's lift-to-drag ratio. The company says a release from 80,000 feet can yield a silent glide of hundreds of kilometers, with a small radar cross-section and no propulsion signature.

Two other design choices stand out:

GPS-denied navigation. Because jamming routinely degrades GPS in contested airspace, Piranha pairs inertial guidance with terrain contour matching (TERCOM) and downward-looking optical navigation, comparing camera imagery against stored digital elevation models. Aventra claims meter-level terminal accuracy without GPS or radio links.
Swappable payload bay. The nose section accepts modular payloads: electro-optical/infrared and synthetic aperture radar sensors for reconnaissance, RF jammers for stand-in electronic attack, or a high-explosive warhead that turns the glider into a long-range loitering munition.

What the flight testing shows

Test record and cost figures as reported by the company; not independently verified. Original diagram.

According to the company, Piranha has flown more than 400 test flights, including during the U.S. Army's Arcane Thunder 26 exercise, run by the 2nd Multi-Domain Task Force, and in multinational special-operations evaluations. Those tests reportedly demonstrated stratospheric release, long-range unpowered glide, and precision delivery under electronic warfare conditions.

The cost claims are the headline: Aventra says a Piranha payload costs about 1/20th of a conventional cruise missile, cutting the cost of high-altitude precision delivery by up to 90%. These figures come from the company and its backers and have not been independently verified.

Open engineering questions

The engineering tradeoffs discussed in this section. Original diagram.

The stratosphere is a hostile place for electronics. Temperatures drop to around -60°C, which degrades lithium battery performance and servo actuators. Keeping payloads warm through passive insulation and high-density batteries — without adding weight that kills glide range — remains an unsolved tradeoff at scale.

Wind shear is the other hazard. Crossing the tropopause exposes a lightweight airframe to abrupt, high-velocity wind changes, and the structure has to survive it on every flight. There is also the simpler question of positioning: a balloon drifts with the wind on the way up, so getting payloads to a usable release point depends on forecasting and launch timing.

Aventra says it will use the new capital to build out production in Herndon to meet demand from U.S. and allied forces.

The bigger picture

The tradeoff between orbital assets and balloon-launched gliders as framed in the article. Original diagram.

Piranha is best understood as a tactical alternative to low Earth orbit. Satellites offer persistent global coverage but cost a lot, take years to launch, and cannot dwell over one battlefield. A balloon-launched glider offers the opposite trade: cheap, on-demand, and local — and disposable if intercepted.

Whether that trade wins real procurement contracts is the open question. The Booz Allen investment and the Arcane Thunder 26 flights suggest the Pentagon is interested. But interest in exercises is not a program of record, and the hard physics of cold batteries and wind shear still sits between a promising demo and a fielded system.