Lab-grown neurons gave a video model a 156,000-parameter edge

Two stories today: a Baltimore startup turned recordings of living brain cells into a shippable AWS product, and Chinese labs got a seat at the UN Security Council's AI table.

The Biological Computing Company is now selling its neuron-derived optimizer through AWS — and the neurons are the least commercial part of it. TBC grows cortical cells on a silicon chip with 4,096 electrodes, feeds them image data as electrical pulses, records how the excitation spreads across the culture and fades, then turns that measurement into a 156,000-parameter software adapter that adds under 0.1 percent to the size of the base video model. On the open-source Oasis model, the adapter held object edges and spatial relationships longer: about 19 percent better than the base model on image differential entropy, roughly 15 percent better than a same-sized conventional fine-tuning and 5 percent better than a LoRA. Amazon is pitching the result as five times faster generation at 80 percent lower inference cost — which, as The Decoder points out, is one effect stated in two units, since needing a fifth of the GPU time is an 80 percent saving.

The commercial picture is thinner than the headline. WIRED reports TBC raised $25 million in the spring, led by Primary Venture Partners, then quietly closed another $25 million — a round not previously reported, taking total funding past $50 million. For the product itself, TBC names neither the base model, the hardware, the resolution, nor the quality metric, and it has published no controlled comparison against purely digital tricks like distillation or caching. Its real thesis isn't that neurons compute better than GPUs. It's that biology surfaces optimization strategies nobody would have found at a desk — a research bet now being sold as a line item on AWS Marketplace, with Trainium and SageMaker deployment planned. AWS's own public caveat is the honest one: push the model to hour-long video and nobody yet knows what fidelity costs.

DeepSeek and Moonshot will brief the UN Security Council on AI risk on Wednesday — the first time Chinese labs share that floor with OpenAI and Anthropic. Reuters reported the invitations; Sam Altman will appear in person, diplomats expect senior Anthropic staff, and DeepSeek founder Liang Wenfeng is not expected to attend. France holds the council's rotating presidency this month and called the session, with a concept note flagging malicious use of AI and the risk of losing control of advanced models. UN Secretary-General António Guterres told the General Assembly the dangers can't be ignored; President Trump has argued existing safeguards suffice. Back in September the two governments agreed to their first dedicated AI safety talks — US, China schedule first dedicated AI safety talks of Trump's second term — and this session extends that channel into the UN's most formal venue.

The symbolism is doing most of the work. A Hangzhou lab that rattled US tech stocks in January 2025 now briefs the same fifteen-member body as the labs publicly accusing it of distilling their models, and the council has no enforcement mechanism to offer either side.

What to watch: whether the Chinese labs use the floor to answer Anthropic's distillation allegations directly, and whether the American labs say anything in New York they haven't already said in Washington.

Should frontier labs have a formal seat at the Security Council, or does that just legitimize the race? Tell us in the comments.

Houmo moves compute inside memory — and A20 enters production

Beijing chipmaker Houmo used its partner conference this week to lay out two compute-in-memory routes at once — and to say the first chip built on the newer one is already in production. Houmo's A20, its first chip built on a 3D compute-in-memory architecture, has entered production. The part supports up to 16 chips fully interconnected, claims a 50% improvement in compute utilization over the previous generation, and covers models from 3B to 30B parameters, according to co-founder and product

Kling 4.0 goes live with 30-second native video at 10-bit HDR

Kuaishou's Kling AI has spent two years becoming the video model people actually finish projects with. Its new flagship, Kling 4.0, went fully live on October 15 — and this round is less about prettier frames than about the two things that make a generated shot usable in a real edit: length and control. Kling 4.0 generates up to 30 seconds in a single native pass, outputs 10-bit HDR at 4K, 1080p and 21:9, and takes up to 15 reference assets in one task. That budget splits across ten images, fi

Brain organoids classified Braille at 83% — in a trio, not alone

Two stories today: human brain tissue on a chip decoded a tactile signal, and the data-centre company behind some of the largest AI contracts in Europe turns out to be six months behind on its flagship site. Human forebrain organoids grown on a microelectrode array classified Braille letters with 61 percent accuracy on their own and 83 percent when the responses of three organoids were combined, in a peer-reviewed study from the University of Bristol. The team took event-based tactile readings

Suleyman says Microsoft's base model beat OpenAI's on a quarter compute

Microsoft finished pre-training its biggest in-house model yet — then used it to make a point about the company it bankrolls. Microsoft AI completed pre-training of MAI-Base-2, a mixture-of-experts foundation model with 80 billion active and 2.8 trillion total parameters, and says the recipe is 4.2–12.3x more pre-training FLOPs-efficient than DeepSeek-V4-Pro-Base. The run itself is the number worth keeping: 16,000 Nvidia GB200 GPUs for 38 days across August and September, at 21% average model

Read more

Houmo moves compute inside memory — and A20 enters production

Kling 4.0 goes live with 30-second native video at 10-bit HDR

Brain organoids classified Braille at 83% — in a trio, not alone

Suleyman says Microsoft's base model beat OpenAI's on a quarter compute