Imagine a ceiling lamp streaming a movie to your laptop. Its LED changes brightness too quickly for your eyes to notice, while a small receiver in your computer translates the pattern into data. That is LiFi: a wireless connection carried by light.

PureLiFi spent more than a decade turning that idea into products for phones, offices and the military.

On August 31, 2026, the company ceased trading and entered administration, making all 42 employees redundant.

Its administrator said a costly shift from licensing the technology to funding and selling its own hardware exhausted the company's cash before it reached profitability.Pallet built flat-pack sleeping cabins that cities could assemble in under an hour and group into staffed shelter villages.

On September 4, 2026, the company ceased operations after nine years. Pallet's own figures show annual village openings falling from 35 in 2021 to five in 2026.

What Was pureLiFi:

WiFi carries data on radio waves. LiFi encodes it in rapid changes in light intensity. A transmitter in a lamp or access point sends those changes across a room, and an optical receiver in a laptop, phone or adapter reads them.

LiFi creates a new wireless lane without adding traffic to crowded radio spectrum. However, light cannot pass through walls, and a person or object can block its path. Covering a whole house therefore requires equipment in every room and compatible receivers in every device.

That being said, walls gave LiFi one of its strongest military advantages. Radio signals can spread beyond a command post, where they may be detected or jammed. LiFi stays inside a defined cone of light, making the network easier to contain.

pureLiFi had two ways to sell this technology: complete networks for specialist customers, or licensed inventions and components for manufacturers to build into their own products.

This is how defense became pureLiFi's clearest market. Kitefin was a portable system with LiFi access points and compatible endpoint equipment for command posts. Because pureLiFi supplied the complete setup, military buyers did not need to wait for existing phones or buildings to support LiFi. In 2021, the US Army placed a reported $4.2 million order for thousands of units.

However, the larger opportunity was putting LiFi inside everyday devices. pureLiFi developed small “Light Antenna” components that phone and laptop manufacturers could integrate into their products. It planned to earn money from licensed technology, components and engineering work as those partners shipped LiFi at scale.

That plan created a loop: A room needed access points and a device needed an optical receiver. Phone makers were waiting for compatible networks, while building owners were waiting for compatible phones and laptops.

On top of that, different suppliers' products could not work together. An official WiFi-family standard for light communications only arrived in November 2023, around eleven years after pureLiFi began commercializing the technology.

To avoid waiting for every device maker, pureLiFi also sold complete products for telecom operators. Bridge XC solved a specific installation problem: an outdoor 5G receiver could get a strong signal, but the connection still had to cross a window or wall. A box outside sent the data through the glass to a matching box inside, avoiding drilling and cabling.

This approach gave customers a finished system. However, it also left pureLiFi responsible for components, assembly, inventory, certification and support. In 2025, telecom operator du was still testing Bridge XC for adoption readiness, and major carriers were running consumer field tests in January 2026.

The Numbers:

💡 Commercial launch: 2012
💰 Capital raised: More than £35M
🪖 US Army order: $4.2M
💷 Cash: £9.44M in 2022 to £2.00M in 2024
🧾 Borrowings: £28.67M in 2022 to £35.81M in 2024
👥 Jobs lost: 42

Reasons for Failure:

Hardware costs arrived before repeatable orders: Licensing would have left more manufacturing and inventory risk with partners. Complete systems required pureLiFi to fund product development, components, testing, delivery and support itself. The administrator said this strategic shift cost more than forecast, while public filings show inventory and equipment rising as the company expanded its product range.
Mass adoption needed an industry to move at once: Device manufacturers, lighting companies, carriers and building owners all needed compatible hardware. Standards and trials showed progress, but neither created volume orders automatically. A large defense purchase proved that LiFi solved a valuable specialist problem; it did not establish how many consumers would pay for another wireless interface.
The financing clock expired first: Cash fell by almost 80% between 2022 and 2024 as borrowings climbed. The 2024 accounts showed £18.15 million of net current liabilities and warned that continued operation depended on more funding. Losses had narrowed from £7.08 million to £4.37 million, suggesting progress, but pureLiFi still needed investors to finance a commercialization timeline controlled by much larger partners.

Why It Matters:

Map every company that must act before the user gets value. The slowest external dependency determines how long the startup must finance itself.
Selling complete hardware can serve customers who cannot wait for partners. Make that move only when committed orders can support the added inventory, certification and working-capital cycle.

Trend

Meta Gets Personal

OpenClaw, Grok Bot, Gemini Spark and Instinct AI are all examples of persistent and proactive personal AI assistants.

Last week, Meta joined them with Muse. Seven days later, the app had reached No. 2 on the US iPhone chart, Meta’s share price had jumped, and Muse looked like the most polished mass-market product in the category.

Why it Matters

Muse gives Meta a different route into the AI race. The company can compete through consumer distribution, identity, commerce and product design without relying entirely on having the best model.
It makes personal agents harder for the major labs to treat as a side project. OpenAI and Anthropic already have much of the machinery in Work, Cowork, Codex and Claude Code. Muse increases the pressure to turn it into something that stays with one person and acts continuously on their behalf.
The category needs its own scoreboard. Successful recurring tasks, human interventions, silent failures, security incidents and cost per completed errand matter more than benchmark scores.

What Muse Actually Does

Muse is an AI assistant that can use a browser, files and a terminal on a remote computer. It can research products, compare options, fill out forms, manage files, run code and complete tasks that require multiple steps.

It remembers your preferences, keeps working after you close the app, splits complex jobs into side chats and runs recurring Goals. You can ask it to research a purchase today, monitor the price next week and buy it once a condition is met.

— (@)

Muse is playing in the same category as OpenClaw or Gemini Spark: personal, proactive and persistent AI assistants that work “in the background”.

— (@)

Meta’s Way Back

For much of the AI race, Meta was visibly behind. Llama had spent a long stretch outside the frontier while OpenAI, Anthropic and Google set the pace.

Muse gives Meta a tangential route back. Personal agents reward distribution, capital, integrations and consumer-product design, all areas where Meta is much stronger than its recent model rankings suggested.

The response was immediate. Muse reached No. 2 on the US iPhone chart and held a 4.9 rating from roughly 8,900 reviews by September 16. META rose 6.55% in the first session after launch and finished the following week 9.25% above its launch-day close.

The Trend

Persistent personal agents are becoming a serious consumer product category. OpenClaw demonstrated the behavior among technical users, Spark brought in Google, and Muse has now packaged the idea for a much broader public.

OpenAI and Anthropic already have agents that use computers, work in the background and complete long jobs. Extending those products into personal agents is an obvious next move. The difficult work sits around the model: memory, permissions, identity, payments, error recovery.

Those problems determine the real benchmarks. A chatbot can lose credibility with a bad answer. A personal agent can send the email, buy the wrong product or stop monitoring the price without warning. Muse has already silently abandoned one monitoring task, suffered repeated logouts and misread uploaded workout data.

Muse looks like the most complete version of this product for now, but reliability will decide the category. The winner will be the agent people trust enough to stop checking every five minutes.

Help Me Improve Failory

How useful did you find today’s newsletter?

Your feedback helps me make future issues more relevant and valuable.

That's all for today’s edition.

Cheers,

Nico