Theme

Mapping and valuing cognitive architecture

Explore this theme’s evidence over time →

Sources

Attribution and reporting linked by this record; links do not imply independent verification.

Welcome to December 12, 2025 (theinnermostloop.substack.com)

Attention turns to the structure of minds themselves — decoding the neural syntax of consciousness, and treating non-standard human cognition as a competitive asset.

Development timeline

Newest reporting dates first. Reporting dates may differ from event dates.

2026-09-20: Stanford researchers propose a dual organization of the brain

2026-09-13: A paper argues a 140,000-neuron brain is the proof of concept for everything larger

2026-09-13: A looped transformer trades parameters for iterations, leaking past compute caps

2026-09-12: A newly mapped fly brain is playing Doom within days

2026-09-04: The first complete male fly connectome maps 166,700 neurons

2026-08-25: A dry-electrode headset decodes silently read words with no sign of saturation

2026-08-13: Twenty-nine of thirty neurotech firms claim unlimited access to brain data

2026-08-10: Researchers reconstruct melodies people merely imagined

2026-08-04: A headstage runs models on-device so feedback never leaves the skull

2026-07-29: Trained volunteers wire a learned category straight to motor output

2026-07-24: The first brain interface certified to restore detailed vision

2026-07-19: Research suggests some frontier models are a small transformer looped twice

2026-07-18: An EEG headset pilots humanoid robots by thought alone

2026-07-12: A researcher hints the next scaling law is parallel test-time compute

2026-07-10: Researchers evolve videos that drive chosen brain regions

2026-07-08: Superposition is read as a mind under pressure

2026-07-08: A model reads silent speech from ultrasound of the tongue

2026-07-08: Taking compression literally turns the timeline into a design question

2026-07-08: A researcher begins eulogizing the transformer

2026-07-05: A pioneer argues a model should learn to model itself

2026-07-01: Sentences are read from brain recordings without surgery at 61% accuracy

2026-06-27: Bilingual brains are found to have no translation neurons, just shared meaning

2026-06-26: A living brain's vasculature is imaged through the intact skull

2026-06-26: Images are generated on a simulated lattice of coupled oscillators

2026-06-25: A founder says brain-to-brain telepathy may be attempted this year

2026-06-21: A robotics lead declares vision-language-action models dead

2026-06-15: A researcher pegs the perfect model at 10.5 quadrillion parameters

2026-05-26: Researchers argue language models need sleep

2026-05-18: Masking an agent's own past actions stops it mistaking hallucination for memory

2026-05-16: An open world model turns one image into a minute of controllable video

2026-05-16: A compact memory state is grafted onto a frozen model

2026-05-15: One module proposes embeddings and another solves for the fixed point

2026-05-14: Averaging token embeddings gives a threefold pretraining speedup

2026-05-12: Models process audio, video and text as a single real-time stream

2026-05-11: Plant seeds are found to wake from dormancy to the vibration of raindrops

2026-05-06: A twelve-million-token context claims a thousandth the compute

2026-05-04: Whole-organism mapping finds vessels space-filling and nerves sheet-like

2026-04-30: A probe puts the frontier model near ten trillion parameters

2026-04-29: A model trained only on pre-1931 text is used to measure the present

2026-04-17: Sperm whale calls are found to pattern like human vowels

2026-04-13: Meta proposes a machine that learns its own runtime from screen traces

2026-04-12: A small quantum computer shows broad advantage for machine learning

2026-04-09: A 4B model dominates long context by rewriting its own weights in flight

2026-04-05: A model improves by fine-tuning on its own samples

2026-04-01: Thirteen parameters reach 91% on a reasoning benchmark

2026-03-29: A compression competition reaches 42.7x over baseline

2026-03-28: Reasoning models turn out to simulate internal debates

2026-03-27: A model trained on 1,000 hours of fMRI predicts brain responses

2026-03-25: A KV cache is quantized to three bits with no accuracy loss

2026-03-22: An architecture scales to a hundred million tokens with under 9% loss

2026-03-20: A lab chasing Solomonoff induction gets tenfold data efficiency

2026-03-18: DeepMind decomposes general intelligence into ten faculties

2026-03-16: Million-token context ships across a model family

2026-03-16: A WebAssembly interpreter is hard-coded into transformer weights

2026-03-13: Data centers are built from lab-grown human neurons

2026-03-07: A whole-brain emulation drives a physics-simulated body through multiple behaviors

2026-03-07: A whole-brain computational model predicted motor behavior at 95% accuracy

2026-03-05: Geometric signs on 40,000-year-old figurines rival protocuneiform

2026-03-05: Parallelizing drafting and verification gives 5x faster inference

2026-03-02: The smallest adder falls from 121 parameters to 36 in a week

2026-02-28: Living human brain cells learn to play DOOM in a week

2026-02-27: Sakana compiles documents directly into model weights

2026-02-25: A 121-parameter adder is hand-coded rather than trained

2026-02-25: A 35B model beats its own 235B predecessor

2026-02-24: A model passes the car wash test

2026-02-24: Anthropic says the assistant is a character the model learned to play

2026-02-20: Newborn chickens show the bouba-kiki effect

2026-02-19: Scaling laws are derived from the statistics of language itself

2026-02-12: Agents meta-learn their own memory architecture

2026-02-10: Musk predicts models will skip source code entirely

2026-02-08: AlphaEvolve discovers an activation function that triples ReLU

2026-02-07: Bonobos are found to identify pretend objects

2026-01-26: Tao wonders whether humans are also predicting the next word

2026-01-24: The speedrun breaks 100 seconds with fewer tokens than parameters

2026-01-15: DeepSeek finds a U-shaped law for memory against computation

2026-01-11: Sandia maps the motor cortex onto neuromorphic silicon

2026-01-11: Model weights are encoded in radio waves

2026-01-09: Some dogs learn words like toddlers

2026-01-08: The speedrun record falls to 109.2 seconds on a new attention mechanism

2026-01-05: Models know they are right before running the code

2026-01-04: Gemini 3 Pro reasons two hops without thinking out loud

2026-01-03: Reasoning as diffusion beats GPT-5 on logic

2026-01-03: Princeton reframes the residual stream as a geometric flow

2026-01-02: DeepSeek widens the residual stream for free

2026-01-02: A model takes over its own context via a Python REPL

2026-01-01: A 40B looped transformer claims 81.4% on SWE-bench Verified

2026-01-01: Adobe formalizes discovering concepts from raw experience

2025-12-31: Transformers are verified to implement Bayesian inference geometrically

2025-12-30: KAIST extracts the prefrontal mechanism for separating goals from doubt

2025-12-30: SmartEM accelerates connectome imaging sevenfold

2025-12-30: Test-time training gives continual learning constant latency

2025-12-29: Diffusion models generate before they memorize

2025-12-29: Pigeons hear magnetic fields through a vestibular circuit

2025-12-29: A System 3 outer loop cuts reasoning steps 80%

2025-12-29: Humans and transformers share learning dynamics

2025-12-17: NeuroBionics reaches the brain through blood vessels

2025-12-15: Mental health disorders resolve into five genetic clusters

2025-12-12: Astera funds a $600M decade on the neural syntax of consciousness

2025-12-12: Palantir recruits explicitly for non-linear thinking

Knowledge graphopen in browser →

Cite or correct this record

Citation formatCitation text

Mapping and valuing cognitive architecture
Innermost Loop KB — secondary record
Reporting date: unknown
KB permalink: https://nicholsn.github.io/innermost-loop-kb/themes/architecture-of-mind
Attributed sources:
- Welcome to December 12, 2025 — https://theinnermostloop.substack.com/p/welcome-to-december-12-2025
This citation does not imply independent verification of linked-source claims.

Suggest a correction — opens a GitHub draft for you to review and submit.