▲ aaron fas

Apprentice of everything, time obsessed, programming lover and a little bit geek. Code Wrangler at @Automattic. Former @TuentiEng ❤️

[▲ aaron fas @aaronfc]

Underrated AI win: not dealing with merge conflicts by hand anymore.

It used to mean understanding some external changes and their motivation, merging them, then hoping a test caught the line we accidentally dropped.

Damn, I don't miss that.

[▲ aaron fas @aaronfc]

@kunchenguid I would go one step deeper: isn't that dangerous for human progress? At a personal level I wrote this yesterday https://t.co/UgLsY6b5b0 I would love to hear about a more positive view haha

[▲ aaron fas @aaronfc]

@poteto I want to believe because it makes sense to me, but reality is that we find so many things at review time that it's scary to just stop doing them. May you elaborate on empirical and formal verification? Is that automated testing coverage?

[m0b1 @mubarak_marafa]

you can just build things. challenged myself to complete my portfolio start to finish in 3 hours. this is the result.

welcome to my corner of the internet.

https://t.co/xTERIu2Dr0

[▲ aaron fas @aaronfc]

@nonamenosocks This is beautiful. Thanks for sharing!

[▲ aaron fas @aaronfc]

@artificialguybr I wouldn't say worse, but disappointingly similar. It's absurdly cheap, though.

[apolinario (poli) @multimodalart]

Introducing the Decision Index 0.1 ⚖️

a rigorous leaderboard comparing jev with 30+ open weights decision models

35+ benchmarks. asking 130K questions to each model

testing knowledge 🧠, automation ⚙️, understanding 🤔and even creativity 🎨

https://t.co/Su4OXqJwz0

[Kun Chen @kunchenguid]

my feedback is that you should allow subscription oauth in 3p harnesses, like almost every other LLM provider does already

claude code is great - don’t get me wrong! but it’s strongly coupled with the claude model

there are many people like me who use multiple LLMs and prefer to work in a vendor-neutral harness where we can keep the same workflow stable across a wide range of models

currently being tied to the CC harness is actually a negative point when it comes to deciding whether to keep the anthropic subscription. not because CC isn’t good, but because it’s inherently vendor-biased and fights against a beautiful multi-model reality

[▲ aaron fas @aaronfc]

@mattpocockuk Isn't this similar to going back to /plan? I really like from /grill-me that it forces you to go one by one. If I get a bunch of questions I find cumbersome having to quote and answer each of them. I would still need to say "I agree".

[Maxime Rivest 🧙‍♂️🦙🐧 @MaximeRivest]

Fable turned my remarkable into Tom Riddle's diary from Harry Potter.

My prompts fade, a LLM respond.

Magical!

[▲ aaron fas @aaronfc]

@mattpocockuk Would you mind elaborating more on what do you mean with "kicking off cross-repo tasks" ? Is it just asking to kickoff work as subagents?

[Artificial Analysis @ArtificialAnlys]

Z ai’s GLM-5.2 is the new leading open weights model on the Artificial Analysis Intelligence Index scoring 51 and it sits on the Pareto frontier of Intelligence vs Cost per Task

@Zai_org’s GLM-5.2 is the same size as GLM-5.1 (744B total / 40B active parameters) but scores 11 points higher on the Intelligence Index v4.1, placing ahead of MiniMax-M3 (44) and DeepSeek V4 Pro (max, 44). On the first-party API it is priced in line with GLM-5.1 at $1.4/$4.4/$0.26 per 1M input/output/cache hit tokens

Key results:

➤ GLM-5.2 is the leading open weights model on the Intelligence Index v4.1. At 51, it leads MiniMax-M3 (44), DeepSeek V4 Pro (max, 44) and Kimi K2.6 (43)

➤ Improvements across most evaluations, particularly scientific reasoning: GLM-5.2 gains over GLM-5.1 on most evaluations, led by scientific reasoning on CritPt (+16 points to 21%) and HLE (+12 points to 40%), alongside AA-LCR (+9 points to 71%), tau3 banking (+15 points to 27%) and SciCode (+7 points to 50%). TerminalBench v2.1 also improves (+16 points to 78%) and GPQA Diamond gains 3 points to 89%

➤ Leading open weights model on GDPval-AA v2 and competitive with proprietary models: GLM-5.2 scores 1524 on GDPval-AA v2, ahead of MiniMax-M3 (1418) and DeepSeek V4 Pro (max, 1328). This impressive result places GLM-5.2 in-line with proprietary models including GPT-5.5 (xhigh reasoning). GDPval-AA v2 builds on the original GDPval-AA by baselining Elo to human performance at 1000, introducing a rotating panel of frontier-model judges, and raising the turn limit from 100 to 250 for longer-horizon agent trajectories

➤ GLM-5.2 uses more output tokens per task than other leading open weights models: the model uses 43k output tokens per Intelligence Index task, up from GLM-5.1 (26k) and above MiniMax-M3 (24k), Kimi K2.6 (35k) and DeepSeek V4 Pro (max, 37k)

➤ On the Intelligence vs. Cost per Task Pareto Frontier: GLM-5.2 is on the Pareto frontier of the Intelligence vs Cost per Task chart, with the lowest cost per task among models at its intelligence level. GLM-5.2 costs ~$0.46 per task, compared to GLM-5.1 ($0.25), Kimi K2.6 ($0.31), MiniMax-M3 ($0.18) and DeepSeek V4 Pro (max, $0.05)

Additional Model Details:

➤ License: MIT

➤ Size: 744B total parameters, 40B active parameters, equivalent to GLM-5.1

➤ Context window: 1M tokens, up from 200K on GLM-5.1

➤ Pricing: $1.4/$0.26/$4.4 per 1M input/cache hit/output tokens

➤ Availability: Alongside Z ai's first-party API, GLM-5.2 is available across third-party providers including @DeepInfra, @novita_labs, @nebiusai, @parasailnetwork , @SiliconFlowAI , @gmi_cloud , @Baseten and @FireworksAI_HQ

[Karthik Hariharan @hkarthik]

“We used to review every line of code before it went into production”.

[▲ aaron fas @aaronfc]

An updated https://t.co/IBalQ1BPa7

[▲ aaron fas @aaronfc]

@steipete @willcb You just had to ask your Openclaw to do it, @steipete!

[binji @binji_x]

The boy will grow up asking the machine before he asks his father. And the machine will be smarter than his father and kinder than his father and more patient than his father. And the father will feel useless and he will be right to feel it. But the boy will still need something the machine cannot give and by the time he knows this he will have forgotten how to ask a person for anything.

[▲ aaron fas @aaronfc]

We are not going back.

I was doing the dishes and realized AI has completely changed how I code. And I don't know exactly when it happened.

First post in a series of thoughts on AI.

https://t.co/ZY5WwFQKZn

#AI

[Andrej Karpathy @karpathy]

I've never felt this much behind as a programmer. The profession is being dramatically refactored as the bits contributed by the programmer are increasingly sparse and between. I have a sense that I could be 10X more powerful if I just properly string together what has become available over the last ~year and a failure to claim the boost feels decidedly like skill issue. There's a new programmable layer of abstraction to master (in addition to the usual layers below) involving agents, subagents, their prompts, contexts, memory, modes, permissions, tools, plugins, skills, hooks, MCP, LSP, slash commands, workflows, IDE integrations, and a need to build an all-encompassing mental model for strengths and pitfalls of fundamentally stochastic, fallible, unintelligible and changing entities suddenly intermingled with what used to be good old fashioned engineering. Clearly some powerful alien tool was handed around except it comes with no manual and everyone has to figure out how to hold it and operate it, while the resulting magnitude 9 earthquake is rocking the profession. Roll up your sleeves to not fall behind.

[Paul Copplestone - e/postgres @kiwicopple]

the @gravatar team just shipped a Wrapper so that you can query gravatar images for all your supabase users

SELECT *
FROM gravatar.profiles
WHERE email = '[email protected]';

enjoy ↓

[▲ aaron fas @aaronfc]

@alain_invasion @gravatar Hey, sorry I missed this answer from yours. No, there is nothing you did wrong. It was a bug due to a recent change we made. Thanks for reporting again!