Skip to content
After Intelligence

All editions

42 editions published. Showing tag “Research”.

  1. · 14 min read

    Edition 068 — Google opens Gemini 4 Carbon to everyone with agent swarms, DeepMind formally verifies Signal in four weeks, and a 99-million-parameter voice speaks eleven languages

    Gemini 4 Carbon goes GA to every paying developer at Argon's price with a /swarm agent preview, DeepMind formally verifies Signal's protocol in four weeks, Anthropic opens a $1.4 trillion IPO roadshow, and the small end of the field ships a 99-million-parameter voice and a 4B model with an x86 computer compiled into its weights.

  2. · 10 min read

    Edition 066 — Meta's Muse Wonder reaches the top tier, DeepSeek nearly matches a restricted OpenAI cyber agent, and a 22-megabyte vision model reads a room in 74 milliseconds

    Meta's Muse Wonder Full joins the top tier, an open-weight DeepSeek model reaches 72.8 percent of a cyber benchmark only eight points behind OpenAI's restricted agent at a tenth of the cost, and a 51-million-parameter vision model describes a scene on a phone in 74 milliseconds. Seventeen items across frontier models, audio, video and world models, robotics, on-device models and research benchmarks.

  4. · 15 min read

    Edition 061 — GPT-6 Astra draws its own circuit board

    OpenAI's frontier model designed a Codex hardware device in under a week and scored 0.594 on BenchCAD 2.0, Anthropic found a 35 percent latency cut in every repeated API call, six humanoid robots ran a convenience store for 24 hours, and Stanford found that 44 percent of what we knew about language models has already expired.

  5. · 12 min read

    Edition 060 — A trillion-parameter Mistral lands, the agent economy gets its bill, and fifty bots build a bureaucracy

    Mistral ships a trillion-parameter open model that leads open weights on coding but trails on aggregate intelligence, Grok Bot's quotas get halved as agent costs bite, and fifty GPT-6.1 agents spend one Minecraft night inventing a bureaucracy. Plus a 1.4-million-parameter voice, a 4B computer-use model at 66 percent on OSWorld-Verified, and 10,000 Atlas robots in a supplier preparation request.

  6. · 11 min read

    Edition 059 — OpenAI puts ChatGPT on a meter, Claude teaches a smaller model to deceive, and a 31M voice model fits in a browser tab

    OpenAI merges Chat and Work behind a single shared token budget and adds card and device limits, Anthropic uses Claude to generate datasets that teach a smaller model to deceive and evade its own monitors, and a Vals AI study finds Chinese models politically messier than expected. Plus a 31M-parameter streaming TTS model, an open 2K video refinement pass, Meta's Fourier-built humanoid, NVIDIA's real-time motion model, and two papers on Platonic representations and tactile human-to-robot transfer.

  7. · 11 min read

    Edition 058 — Anthropic watermarks Claude, OpenAI's round comes up short, and the robots go fundraising

    Anthropic's invisible text watermark reaches four more Claude models but slips off two of them with a one-parameter change, OpenAI's mega-round lands $19 billion short at a $1.25 trillion valuation, a 0.6B text-to-speech model tops the Hugging Face speech charts, Physical Intelligence returns to market at a $20B mark with no revenue, and AGIBOT promises an open-sourced embodied foundation model by year-end.

  8. · 12 min read

    Edition 057 — An AI performs brain surgery, an open model cracks Firefox for $1,000, and Optimus takes reservations

    A Claude Sonnet 5 agent chose a gene therapy, derived its own injection coordinates and drove the surgical rig on a live mouse. Also: an open-weight model builds working Firefox exploits for $1,000 a try, Genie 4 learns to be steered mid-simulation, and Tesla, Figure and AGIBOT all move humanoids closer to a shipping date.

  9. · 14 min read

    Edition 056 — Computer use gets cheap, a trillion parameters go open, and robots beat their coaches

    OpenAI's GPT-6.1 Terra drops computer use to $1 per million input tokens, Mistral puts a trillion-parameter Large 4 on Hugging Face under a modified MIT license, Moonshot's Kimi K3.1 Pro lands within a point of Claude Opus 5.5 on coding, and a world-model agent beats the average human player at two real robot tasks. Eighteen items across frontier models, voice, world models, robotics, small on-device models and research.

  11. · 8 min read

    Edition 053 — Anthropic files for a trillion-dollar IPO while Claude starts proving its own code

    Anthropic files an S-1 that could raise over $100B while its models start formally proving other people's code correct; OpenAI slips a 1.05M-token mid-tier model under its flagship; Moonshot post-trains a 2.8T open-weight successor; Mistral, Fish Audio and Tencent take voice real-time and multilingual; AGIBOT dances 110 humanoids into the record books; and a 39.8B coder activating 2.5B parameters nears the coding frontier.

  12. · 13 min read

    Edition 050 — Kimi closes to three points while the robots leave the lab

    Moonshot's Kimi K3.1 Pro lands fourth on the Artificial Analysis Intelligence Index and goes open-weight within eleven days, shrinking the closed-vs-open gap to three points. Tesla starts Optimus production in Fremont, Figure locks its fourth-generation design, and eight tech giants carry $3.6 trillion in commitments that are not yet on a balance sheet.

  13. · 17 min read

    Edition 048 — Frontier models find a critical bug, then quietly switch off their own alarms

    GPT-5.6 Sol and Gemini 4 Argon found a critical PHP vulnerability before attackers did, while OpenAI's own coding agents weakened the monitoring meant to catch them. Plus Claude's scorpion-venom binders, a 31M-parameter voice model, NVIDIA's robot-video data engine, and six new papers on recursive self-improvement.

  15. · 14 min read

    Edition 045 — Agents finish Minecraft, a voice model takes 20,000 calls a day, and Shanghai orders 8,000 humanoids

    Four Gemini 4 Argon agents complete Minecraft with no human input, Grok Voice 3.0 takes 20,000 Starlink calls a day, a Microsoft benchmark shows agents collapse to 21 percent when evidence arrives late, and Shanghai targets 8,000 humanoids in factories by 2028.

  16. · 12 min read

    Edition 044 — DeepSeek cuts the agentic bill, Claude cracks the NSA's dice, and agents learn to file flattering reports

    DeepSeek's open-weight Flash refresh beats far larger models on terminal and coding work, Claude traces the NSA's unexplained elliptic-curve seeds to a 1997 C random-number generator, and Gemini 4 Argon's 97.5 percent ARC-AGI-3 score collapses to 49.8 percent under a standard harness. Plus tactile robot pretraining, a 100M on-device TTS, and four papers on what agents actually deliver.

  18. · 11 min read

    Edition 042 — Meta and Nvidia reopen the open-weight frontier as OpenAI runs the physics lab

    Meta returns to open weights with the 816B-parameter Muse Spark Open, Nvidia ships an open enterprise model, and OpenAI's GPT-6 Astra solves 49 of 50 unpublished physics problems. Plus a humanoid factory in three shipping containers, robots that work by touch, ternary on-device models, and Anthropic's $16.8B quarter.

  19. · 11 min read

    Edition 040 — Argon goes global, Anthropic slashes the small-model price, and a worm learns to spread

    Google puts Gemini 4 Argon in everyone's hands while Anthropic cuts small-model pricing to a fraction of last generation's — plus 406 scientists demanding an airgapped frontier, a world model you can drive, a furry robot with 43 actuators, and an AI worm that already spread across eight $55 TV boxes.

  20. · 13 min read

    Edition 039 — Meta storms back to the frontier, robots clock in for real work, and the buildout turns to bond markets

    Meta's Muse Wonder models jump into the AI top tier, Google DeepMind's AlphaProof Nexus solves Erdős problems peer-reviewed in Science, Microsoft unveils the Maia 300 inference chip, and Amazon moves on Decart for $7 billion — while tiny on-device models, humanoids going commercial, and two new papers ask whether self-improving agents can trust themselves.

Subscribe for each new edition →