Permanent copy of issue 60. Today's issue is always the front page.
Previous issue | This week | Next issue | archive
Samsung said Tuesday it will put a combined $1 billion into Helix Digital Infrastructure, a company that builds data centers and the power to run them, alongside Nvidia and U.S. power producer Vistra. OpenAI will not release GPT-6.1 Astra after its head of safety systems said the model did not meet the company's bar. India is lining up $25 billion for deep-tech startups, CNBC reported. The machines, and the decision to ship a model, still sit with the cluster.
Date: 2026-09-29
Updated: 2026-09-29 06:04
Status: LIVE
Broadcast
OpenAI said it will not release GPT-6.1 Astra, a model meant to browse the web and use apps on its own, because it did not meet the company's safety bar. The model a household can rent, GPT-6 Astra, still runs in OpenAI's data centers.
Watch
Stories
Samsung said Tuesday it will invest a combined $1 billion in Helix Digital Infrastructure, with Samsung Electronics putting in $500 million and five other affiliates the rest. Yonhap reported that Helix, led by former Amazon Web Services chief Adam Selipsky, was launched in June by KKR, Nvidia, the Kuwait Investment Authority, and U.S. power producer Vistra, with more than $10 billion already committed.
-- The check buys a stake in cluster buildings and power plants, not a chip a household can put in a desktop.
2026-09-29 / Yonhap News Agency / chip, energy / central / #samsung-puts-1-billion-into-helix-data-centers
OpenAI has announced it will not release GPT-6.1 Astra, the BBC reported Tuesday. Saachi Jain, head of safety systems at OpenAI, said the system, which browses the web and uses apps on its own, "didn't quite meet the bar" of the company's standards.
-- The hold is OpenAI's decision. A person at home does not get the model, and does not get the switch that stopped it.
2026-09-29 / BBC / efficiency / central / #openai-shelves-gpt-6-1-astra-after-safety-tests
India is preparing to deploy $25 billion for deep-tech startups, including artificial intelligence and semiconductors, CNBC reported Tuesday. Rajat Tandon, president of the Indian Venture and Alternative Capital Association, told CNBC the government has committed $11 billion through a research fund, to be matched by private capital, with a further $3 billion to $4 billion on top.
-- The pool is aimed at national labs and startups. It does not hand a household a model it can run itself.
2026-09-29 / CNBC / chip / central / #india-lines-up-25-billion-for-deep-tech-startups
Anthropic said Monday that Claude Sonnet 5.5 runs more than 30% faster than Sonnet 5 and costs up to 30% less for most work. The list price is unchanged at $2 per million input tokens and $10 per million output tokens; Anthropic said the saving comes because the model uses fewer tokens.
-- A cheaper task still means renting Anthropic's cluster. The weights stay off the home machine.
2026-09-28 / Anthropic / efficiency / mixed / #anthropic-ships-sonnet-5-5-at-the-same-token-pri
Axis
local [########################------] central
Score: 79
clusters still winning
up 1
Samsung put $1 billion into a Nvidia-backed company that builds data centers and power plants. OpenAI held GPT-6.1 Astra. India is lining up $25 billion for deep tech. The cluster still holds the model.
Forces
| Force | Meter | Score | Pull | Note | ||
| CHIP / CAPEX |
|
91 | central | Samsung's $1 billion and India's planned $25 billion both fund clusters and chip startups. No new chip a household can buy landed overnight. | ||
| ENERGY |
|
96 | central | Helix names U.S. power producer Vistra as its preferred power partner. The new capital is for generation and data centers, not a credit on a home electric bill. | ||
| EFFICIENCY |
|
56 | central | Anthropic left the Sonnet list price at $2 and $10 per million tokens and said a typical task costs less because the model uses fewer tokens. OpenAI held back a more autonomous model. Neither change puts weights on a desk. |
Line charts (0=local 100=central 2026-09-14 to 2026-09-29)
Market overlay (2026-08-31 to 2026-09-29)
Latest: BTC $84,168 / hashrate ~951.04 EH/s / token markers avg $5.3033/M / axis 79. 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 $84,168 / hashrate ~951 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
One email per issue. No tracking pixels, no filler.
home | rss | page 2 | page 3 | archive | models | subscribe | data | board | mcp | llms.txt