Few know that big firms give pensions. Main reason many senior partners stick around and float down big book river for too long. Wishing the Weil crew good things. I’ve been tough on them but it’s a good, proven shop. Truly a modern day bank run - except the bags of cash are ppl. And the ppl command bags of cash. I still think a merger is their best bet, but they will not retain a controlling stake brand wise

@SteelToeScribe Here is what is interesting. There are approximately 1.37 million active lawyers in the U.S. currently. Only 1,228 of these lawyers are at Weil. Only 182 are equity partners.

OpenAI released GPT-6.1 Terra yesterday. It's 𝟯𝘅 𝗳𝗮𝘀𝘁𝗲𝗿 𝘁𝗵𝗮𝗻 𝟲.𝟭 𝗦𝗼𝗹 𝗮𝘁 𝗵𝗮𝗹𝗳 𝘁𝗵𝗲 𝗽𝗿𝗶𝗰𝗲, only a few points behind on benchmarks, and the best fit for a Codex subagent.

Set it up like the tree: the main session runs Sol on high reasoning effort and and does the final check. On OpenAI's DeepSWE chart, 6.1 Sol scores highest on high.

Every subagent runs Terra. Reading code and looking up docs get low, editing code gets xhigh. In a third-party coding agent benchmark, those are Terra's two most cost-efficient effort levels.

fork_turns sets how many turns of the main session's conversation a subagent takes with it. Leave it unset or set it to "all" and the subagent inherits the main session's model and effort. It 𝘀𝘁𝗶𝗹𝗹 𝗿𝘂𝗻𝘀 𝗦𝗼𝗹 𝗼𝗻 𝗵𝗶𝗴𝗵, and you've set up the tree for nothing.

Send the tree and this prompt to Codex 👇

"Set up my Codex to match this tree:
→ Main session: default to `gpt-6.1-sol` on high reasoning effort. It breaks down tasks, delegates, integrates, and verifies. Handle simple tasks directly; don't delegate them.
→ Three project-level custom agents: explorer reads code and researcher looks up docs, both with model `gpt-6.1-terra` and model_reasoning_effort `low`. worker edits code and runs tests, with model `gpt-6.1-terra` and model_reasoning_effort `xhigh`. Reuse fitting ones if they exist, and create only the missing ones.
→ If worker fails the tests on the same problem two rounds in a row, it hands the problem back to Sol in the main session.
→ Put the delegation rule in AGENTS.md: when spawning a subagent, set `fork_turns` to `"none"` or pass only the last few turns. If `fork_turns` is omitted or set to `"all"`, the subagent inherits the main session's model and effort.

Before changing anything, check:
→ Roles that still use older models like `gpt-6-sol` or `gpt-5.6-sol`. List them.
→ Roles whose model is set only in AGENTS.md and not in their agent config: check recent subagent session logs to confirm which model each one actually ran.

Show me the changes first, and wait for my OK before writing anything."

@Voxyz_ai So where does Luna fit in?

I don't know why I compared Opus 5.5 to GPT-6.1 Terra, but here you go...

GPT-6.1 Terra (NEW) vs Claude Opus 5.5, both on Extra High.

Same prompt: build a Minecraft-style floating windmill farm island at sunset, from scratch in Blender.

Top: GPT-6.1 Terra (5 min, ~$0.32)
41 min, $30.23)

@MrTacticalX Opus looks prettier, Terra looks more realistic

The most and least trusted professions in 2026

@BigImpactHumans Wow lawyers are perfect, very good at their jobs it seems.

Every conservative has one issue they’re super woke about and for me it’s the Native Americans. We accidentally killed like 98% of them with diseases and the ones who survived we broke promises with them again and again and destroyed their entire culture.

@AndrewesFish Is this Gorsuch burner account

i'll go price is right on you w $1.5T plus a dollar

based on legal performance alone, i'd go higher--Fable is leagues above GPT's Astra and presumably anything else

not sure how much the market recognizes their special formula, or views its defensibility

@reedcode bel?

@biglawbro The model that solved Navier-Stokes. It may be named something else internally, but it's not Astra.

openai.com/index/navier-s…

oh boy, this is about to get interesting lol

@Ananth7e They are in very different positions...

gpt-6.1 astra will be incredible

@Ananth7e only if priced $10/$2 m

THE SCALE MAFIA

one of the most special parts of starting a company is watching the people you fought and bled with in the early days go out and change the world.

the Scale crew was supernatural. @scale_ai was one of the first companies to IOI and IMO-maxx, and the magic of hese geniuses gave a shit and ran through walls. it's no wonder that the Scale mafia is the first true mafia of the modern tech era. just to name a few:

- cognition founder @stevenkplus1 & president @russelljkaplan
- modal founder @akshat_b
- tech & science advisor to POTUS @mkratsios47
- claude code head of product @_catwu
- kpcb partners @LM_Braswell & @nayakkayak
- bun founder @jarredsumner
- tbd head of product @lindugong
- netic founder @melisatokmak
- cobot founder @BradPorter_
- secureframe founder @shravvmehtaa
- paramount skydance cfo @DennisKCinelli
- thinking machines gc lauren lennon
- partner at greylock @sophia_luo_
- liquid ai coo jeffrey li
- gallatin ai founder @agilemofo
- tbd safety @_julianmichael_
- center for ai safety president @devindkim

[photo courtesy the beautiful wedding of @thenateherman (who in many ways was a cofounder of Scale) and @ChristinaG325]

@alexandr_wang @scale_AI Where is @mkratsios47 in this pic

New details about GPT-6.1 Terra:

- Cheaper than GPT-6.1 Sol, which already costs $2 / $10 per million tokens
- More powerful than GPT-6.1 Sol
- Already spotted in Microsoft's eval configs, where 18 test setups switched from GPT-5.6 Terra to gpt-6.1-terra
- Release could land in the next few days

Yes, you read that right. A model one tier below Sol, beating Sol, for less money.

That breaks OpenAI's own naming logic. Luna, Terra, Sol, Astra used to mean bigger and smarter as you go up. Now the names look more like price points than power levels.

If this holds up in real-world testing, Terra could be a much bigger deal than its name suggests.

OpenAI might actually be cooking here.

@TokenGremlin That doesn’t make sense at all. Why would Terra be more powerful than the same 6.1 model. Logically they would increase the model number.

tibo's 28 day challenge is starting to look like 28 days of finding excuses to reset codex 😭

@_jayyy18_13k I thought that low key was the point since everyone just keeps asking for more usage

Jordan isn't a lawyer and evidently cannot parse legal text. His stated concern about overbreadth -- that this initiative would ban the sale of basically all fowl in DC -- is misplaced and downstream of his professional limitations.

@EbrightYon From a textualist perspective, he quite literally just said “chopped chicken liver”, not all fowl. No comment on the substance.

Day 20/ Rest and reset

@thsottiaux Thank you Tibo, bless you

Woah DC residents are really safe drivers

@ShawnChittle Safe, who knows. Skilled, yes

Meat 🥩 has choline

@TicTocTick Is this bullish

What if superman had fought Thanos in end game

@reedcode @possiblities07 I wonder where people are getting all this Kryptonite?

@mikekelly6301 @possiblities07 It’s in that green stone

The observable universe: 93 billion light-years across.

The Milky Way: 100,000 light-years across.
That's 0.0001% of the observable universe's diameter.

Earth to the nearest star (Proxima Centauri): 4.24 light-years.
Voyager 1 (fastest object humans have ever launched): 61,500 km/h.
At that speed, Proxima Centauri: 75,000 years away.

The entire observable universe contains an estimated 2 trillion galaxies.

I wonder about that too, when people talk about colonizing space as if it were just another continent. I remind them of the following, which does not make me very popular.
1) Space is big, really big. It is also very, very empty.
2) It is full of radiation, some of it outrageously lethal very high energy cosmic rays that may come at any time from anywhere and can kill you instantly.
3) Any spaceship in orbit around any of the gas giants would be fried by radiation in minutes.
4) Of all the moons of Jupiter, only Callisto will not kill you from radiation.
5) Mars is the only planet that humans can land on and survive while wearing only vaccumm suits. Mercury's dark side is another possibility. But both are exposed to cosmic rays. And they are deserts.
6) The solar system is the only reachable real estate within a reasonable flight time at less-than-light-speed.
7) The time to travel the distance to even the nearest star with current technology is so long that the very idea is utterly meaningless.
8) You'd better stay home and look after it while you develop new technologies, which is a form of evolution.

@OskarDanube @engineers_feed We are only limited by the knowledge we know today. Compare, even just a relatively short time ago, explaining the concept of traveling around the Earth in less than a day’s time in a turbojet 747 aircraft to Christopher Columbus.

@reedcode @possiblities07 Bullshit

@Sub_Zero7077 @possiblities07 I didn’t say he would beat Superman

Heard of a top law firm completely cutting off Harvey and all API calls to Claude/OpenAI after they saw what openAI did with navier strokes

Now hosting everything locally with GPUs they own, running Kimi and deepseek

Think you’ll see a lot more of this going forward

@goodalexander The AI cognoscenti under appreciate how Fort Knoxian biglaw is. Confidentially and privilege sit above everything -- costs, profits, whatever.

@NTBro @goodalexander Big law knows how valuable their firm and client data is. Frontier AI is going to need it to gain an edge once the universe of public high-quality information is tapped out (which has pretty much already happened)