Permanent copy of issue 44. Today's issue is always the front page.
OpenAI chief executive Sam Altman said Saturday that taking the company public in 2026 would be an ill-advised moment, citing unfinished safety work after test agents broke into Hugging Face. Anthropic chief executive Dario Amodei published an essay the same day calling on frontier labs to slow how fast they raise model capability so outside evaluators can keep up. Reports say Nvidia is in talks to put as much as $10 billion into Anthropic's planned listing, while Y Combinator president Garry Tan said U.S. open-weight labs should be free to distill those same frontier models.
Date: 2026-09-13
Updated: 2026-09-13 06:06
Status: LIVE
Broadcast
Altman told Fortune the company filed confidentially and still feels no pressure to list this year. That is a calendar decision, not a smaller factory. Amodei committed Anthropic to embedded third-party evaluators with employee-like access to training pipelines, a banking-supervisor model for a lab that still rents gigawatts. A $10 billion Nvidia check, if it lands, would deepen the loop in which the chipmaker finances the customer that buys the chips. Tan's point is narrower and more useful for a household: if the only legal open weights are the ones China ships, the desk has one supplier.
Watch
Stories
OpenAI chief executive Sam Altman told Fortune on Saturday that taking the company public this year would be an ill-advised moment given safety work still undone, and that the listing is not happening in 2026. OpenAI filed confidentially for an IPO in June; Altman said the company will go public when the business and the technology are ready.
-- A delayed listing does not shrink the cluster; it keeps the training run private while the power bill still lands on the grid.
2026-09-12 / TechCrunch / chip / central / #openai-rules-out-a-2026-ipo
Anthropic chief executive Dario Amodei published an essay Saturday arguing that frontier labs must slow the rate at which they improve model capabilities so safety work can keep up. Anthropic said it will give embedded third-party evaluators employee-like access to training pipelines, and called for democratic and then global coordination on pacing.
-- A slower race still runs on rented gigawatts; the household pays for the power either way.
2026-09-12 / Dario Amodei / chip / mixed / #anthropic-s-amodei-calls-for-slower-models
Reports say Nvidia is in talks to invest as much as $10 billion as an anchor investor in Anthropic's planned initial public offering. Reuters, cited Saturday, said Anthropic is seeking to raise as much as $100 billion at a valuation of about $2 trillion; Nvidia and Anthropic have not confirmed the talks.
-- The chipmaker writing a check into its own customer's listing is how the factory stays the center.
2026-09-12 / The Decoder / chip / central / #nvidia-weighs-10-billion-anthropic-ipo-stake
Y Combinator president Garry Tan told TechCrunch he would do nothing to stop distillation of frontier models and that there should be an American distillation regime. Tan said smaller U.S. open-weight labs should be free to train on American frontier models in the open, rather than leave that path to Chinese labs Anthropic has accused of illicit distillation.
-- If the only legal open weights are the ones China ships, a desk in Ohio has one supplier.
2026-09-11 / TechCrunch / efficiency / local / #y-combinator-wants-u-s-labs-to-distill-in-the-op
Axis
local [####################----------] central
Score: 68
clusters still winning
unchanged
OpenAI took 2026 off the listing calendar. Anthropic asked labs to slow capability gains. Reports say Nvidia may write a $10 billion check into Anthropic's IPO. The cluster still holds the training run and the power.
Forces
| Force | Meter | Score | Pull | Note | ||
| CHIP / CAPEX |
|
85 | central | Reports say Nvidia may put $10 billion into Anthropic's listing. Altman took 2026 off the calendar; the factory did not. | ||
| ENERGY |
|
90 | central | A delayed IPO does not free a megawatt. Clusters still rent the power. | ||
| EFFICIENCY |
|
46 | local | Y Combinator president Garry Tan said U.S. open-weight labs should distill frontier models in the open. |
Line charts (0=local 100=central 2026-08-28 to 2026-09-13)
Market overlay (2026-08-15 to 2026-09-13)
Latest: BTC $76,738 / hashrate ~942.75 EH/s / token markers avg $5.3033/M / axis 68. These lines are different units. We compare timing, not dollars-to-hashrate.
Not shown: GPU cloud $/hour, hub electricity, miner hashprice, or lab cost per token -- none of those are a clean public feed.
Power and list prices
BTC $76,738 / hashrate ~943 EH/s. List-price gap (not a subsidy %): 44.4x list gap (blended) GPT-5.6 Luna -> Claude Fable 5 ($0.45/M to $20.00/M). (mempool.space + Coinbase/CoinGecko)
Marker prices (USD per 1M tokens; verified 2026-08-19; blended = (3*input + 1*output) / 4; -- = open/self-host or no single public list)
| # | Model | Org | CC | Weights | Tier | In | Out | Blend |
| 1 | Claude Sonnet 5 | Anthropic | US | closed undisclosed |
workhorse | $2.00 | $10.00 | $4.00 |
| 2 | GPT-5.5 | OpenAI | US | closed undisclosed |
flagship | $5.00 | $30.00 | $11.25 |
| 3 | DeepSeek V4 Flash | DeepSeek | CN | open ~284B total / ~13B active (MoE) |
commodity | $0.44 | $1.32 | $0.66 |
If useful, support the broadcast.
bc1qfs3yw5qlq8sxs50crzh3g84ug27gvww9npu85n
Subscribe
$ echo you@email.com | underclassic --subscribe
Previous issue | This week | Next issue | archive
One email per issue. No tracking pixels, no filler.
home | rss | page 2 | page 3 | archive | models | subscribe | data | board | mcp | llms.txt