https://t.co/ykp1zcNcf0

@andykonwinski I'm trying to help. Built lists of entire AI community here on X: https://t.co/fasUz7PuHq And built an AI to watch it all and pull out the best stuff: https://t.co/kiuZ7QXLzb What else can we do to help?

Elon warned about the AI energy bottleneck before most people even understood it was coming “I can’t emphasize enough - we need more electricity. However much electricity you think you need, it’s more than that” Now if look at what is happening. AI labs are raising billions, buying tons of GPUs, and announcing giant clusters But then reality hits: Where do you plug them in? This is why the AI race is really an energy race. Even the most sophisticated labs can have the money, the chips, and the models but none of it scales without the grid capacity to feed it The bottleneck is not just GPUs anymore. It is electricity, substations, transmission, permitting, and the ability to actually plug these massive clusters into the real world Elon saw this years ago and now everyone else is catching up

New research on beneficial RL: models trained on a small amount of beneficial trait data improve on a wide range of alignment and benefits evaluations, even if trained only on health domain data. We hope it’s a step towards more broadly and persistently beneficial models. 🧵 https://t.co/bN5usoNxCD

Smoke em up Johnny https://t.co/VMXGtrvpeB

six different charts all quietly confirming the run is fine and i am not. credit: @0rdlibrary https://t.co/K9sana594L

six different charts all quietly confirming the run is fine and i am not. credit: @0rdlibrary https://t.co/K9sana594L

“It’s an astonishing story” They’re feeding your children to the crocodiles and the Government won’t say who did it because it would be ‘racist’ to do so… https://t.co/QiqmoIJGAf

@peakcooper Hm, are you sure it’s not a CLAUDE .md in your local folder or perhaps a global one that your harness reads from? https://t.co/3TDHgW5Zcf

Gary the snail 🐌 and the beginning. @boringcompany https://t.co/JrrpnsmGIl

+1 more

We built the fastest PDF -> markdown parser in the world 🚀⚡️ AND it’s more accurate than any other open-source, model-free parser (pymupdf4llm, opendataloader, pdf-inspector, markitdown) on 3 standardized benchmarks: olmOCR0-bench, opendataloader-bench, ParseBench Introducing LiteParse v2.1. The v2 base version was already the fastest document->text parser on the planet, and with this new release we’ve introduced markdown. It is fully open-source (Apache 2.0) and free, is usable from CLI/Rust/Node/Python/WASM, and is also installable as a one-click agent skill. Check it out: https://t.co/7oFImAZeb2 Come check out LiteParse: https://t.co/JNER0mVcB8

+1 more

@rcanand @Abaybektursun Can't confirm https://t.co/XIBo0IfMSZ

Falcon 9 is vertical at pad 4E in California ahead of tomorrow’s launch of the @NRO_gov’s NROL-179 mission → https://t.co/ikCA04Rzvz https://t.co/QdwFj4NRRW

Some moments stay with you. A look back at the extraordinary team behind a milestone achievement. Morgan Stanley was honored to support @SpaceX team through its IPO, serving as joint lead bookrunner and hosting the opening of trading in our offices last week as sole stabilization agent.

+2 more

RILEY GAINES ON BIRTH CONTROL: “So I was on birth control in college, but only because — as a swimmer — you can imagine it’s not exactly comfortable to bleed in a sport like that. The bathing suits are pretty revealing, and we’re in the water six hours a day. “I got on it specifically to stop my period entirely. Looking back now at 26, if I could go back and tell 18- or 19-year-old me anything, I would absolutely tell her not to do that. “I just had a pretty comprehensive health screening — blood work, cancer screenings, the works. One of the only things the doctors flagged was that I had super low bone density. I thought that was strange because I lift weights, which should promote bone growth. So I asked the doctor what could cause it. “She asked if I had been on birth control that let me skip my periods. When I said yes, she told me that was likely the cause. She said it’s good that we caught it now — I’m young and healthy, so I can work to build it back up — but if I had stayed on it longer, it could have really hurt me in the long term.”

twitter is bugging out again https://t.co/ABJGE81e0L

BREAKING: Starlink airline partnerships are growing rapidly. • 2 airlines in 2022 → 41 airlines in Q2 2026 • More than 20× growth in just four years • Contracts now cover 7,000+ aircraft worldwide • Major carriers including United, American, Lufthansa, Air France, Qatar Airways, Emirates, and British Airways are adopting Starlink Starlink is quickly becoming the new standard for inflight connectivity.

SpaceX has just officially received an investment grade credit rating of Baa1 from Moody's, slightly higher than @Tesla's Baa3 rating. Their reasoning: "SpaceX's Baa1 issuer rating reflects the company's exceptional franchise strength as the world's leading orbital launch provider and operator of the largest low earth orbit satellite broadband network, Starlink. The company benefits from robust and expanding recurring revenue from Starlink, which has become the primary cash flow generator and underpins improving scale, margin expansion, and diversification away from more cyclical launch revenues. Additionally, SpaceX's vertical integration across manufacturing, launch, satellite deployment, and end-customer delivery drives superior cost efficiency and operational velocity that competitors have been unable to replicate. Strategic relevance to the U.S. government, as the primary launch provider for NASA and the Department of Defense, adds demand visibility and long-cycle contract support, while the company's ability to monetize AI compute capacity through third-party arrangements provides additional revenue diversification and downside optionality. The AI segment addresses a large total addressable market across AI infrastructure, consumer and enterprise applications, and digital advertising, and represents significant long-term revenue and earnings upside as the company scales its compute infrastructure, develops its frontier models, and expands enterprise and consumer monetization channels. The company maintains conservative financial policies, alongside strong liquidity and financial flexibility, supported by substantial cash balances and continued access to both equity and debt capital markets."

History listed on @Nasdaq. @SpaceX ($SPCX) raised a record $85.7B in capital, hit a $2.1T market cap, and traded 500M+ shares on its first day as a public company. Read more about how the capital injection will directly fund SpaceX’s infrastructure for the future: https://t.co/AZ4uPAtOvZ

TBC is hiring exceptional mechanical, electrical, civil, software, and field engineers in Bastrop, Las Vegas and Nashville. Apply for the July 1 Boring Factory video tour, and hear directly from the engineers who are designing/building Prufrock and trying to Beat The Snail! https://t.co/MolT6mjOka

BREAKING: Paraguay receives Starlink kits to connect remote schools, health centers and communities. 🇵🇾 • 50,000+ students and teachers will benefit. • Goal: connect 1,600 locations nationwide. • Prioritizing rural, and isolated communities. • Also connecting remote healthcare centers. • 100 kits already installed in rural Chaco. • Builds on an initial plan to connect 500 strategic sites. • Expanding access to online education, digital libraries and global opportunities. This partnership between Paraguay Government and Starlink will help bring internet access and more opportunities to people across the country.

The 4-bit quant is now available, making our first open-source agentic coding model small enough to be run on a Mac. Get the weights here: https://t.co/8amCVzzunE

@Abaybektursun Not sure tbh. If yes, it wouldn't surprise me, it's pretty common. Claude distills from the internet & potentially others, others distill from Claude,... that's just the natural dev cycle. GLM 5.2 is >10 pts better than Opus 4.8 on coding tasks btw, so they not "just" distilling https://t.co/ih3FlbLIgB

Where will Midjourney's body scanner improve most dramatically? @iScienceLuvr, CEO of @SophontAI: "This is just a prototype of what can be done, and the resolution will continue to improve over time." "Even just the reconstruction algorithm, you have the ring of all the transducers, and to construct that slice, you have to emit from all different angles and based on that, reconstruct a slice. There are many innovations that could take place in just improving that algorithm." "The algorithm they currently use is a bit of a more basic algorithm. But they're looking at developing all these sophisticated physics simulators to understand the propagation of the ultrasound through the human body and then being able to back-propagate that and construct the slice." "In probably version three of the system, they're gonna be having their own silicon, custom chips they're developing for this. All of that is needed in order to make these advances."

my notes from the @midjourney medical launch - @Scobleizer compared this to the original iPhone and Tesla launches (that he was also front row for) - find you a man who looks at you like @bryan_johnson was 😍 ing for @DavidSHolz - see @iScienceLuvr tweet linked for Nature paper - reminds me of our @biohub episodes: better science starts with better data, and that means better imaging - people asking "but wen FDA?" are so small minded. we will do the easy stuff, then we'll do the harder stuff. roll up your sleeves and help or just be patient. - when you have genuinely better tech+mission, all the other hurdles just sort of fall away/figure themselves out: business model, regulatory approval, hiring, marketing, confusion over what to do - this was just the first of 8 side project launches MJ has planned this year - this is what technological ambition looks like: not 10% better, not 2x better, but 40-100x better in every dimension - how are we getting this level of innovation and ambition out of a $10m/yr research budget and whats wrong with the way we use R&D in every other megacorp/goverment/frontier lab? - how has $BFLY stock not mooned yet, this thing just had its ChatGPT moment thank you to L for letting me into what I believe is going to be the top 10 most important launches i'll ever see live.

Announcing AA-Briefcase, the benchmark for the next era of agentic knowledge work AA-Briefcase is our new benchmark for testing models on long-horizon knowledge work tasks in complex projects built by industry experts. Models are evaluated on multi-week projects, each with many linked tasks and thousands of input source files. We evaluated Claude Fable 5 from @AnthropicAI before it became unavailable, and it currently leads with an Elo score of 1587, followed by Claude Opus 4.8 (max, 1356), Opus 4.7, and the recently-released GLM 5.2 (max, 1266) from @Zai_org. Claude Fable 5 cost $31 on average to run each AA-Briefcase task, followed by Claude Opus 4.8 at $10.40, GPT-5.5 (xhigh) at $3.68 and GLM-5.2 (max) at $2.40. AA-Briefcase comprises four private scenarios, each representing a multi-week knowledge work project set in a realistic organizational context. A public fifth scenario has been released via @huggingface as a representation of scenario structure, submission, and grading (AA-Briefcase Lite). This does not count toward official AA-Briefcase results, and is demonstrative only. Key elements of AA-Briefcase: ➤ Realistic long-horizon projects: AA-Briefcase moves beyond single, disconnected prompts by evaluating models across a coherent long-horizon project. Tasks build week by week, draw on shared institutional context, and require deliverables such as financial models, board presentations, and design mock-ups ➤ Large volumes of fragmented context: AA-Briefcase requires models to reason across thousands of inputs, including company documents, meeting transcripts, large-scale data exports, 25,000+ Slack messages and 3,500+ emails. These sources are fragmented, messy, and often contain realistic contradiction, testing whether models can navigate the ambiguity of real-world knowledge work ➤ Composite rubric and pairwise grading: AA-Briefcase combines binary rubric checks for ground-truth correctness with pairwise grading on analytical quality and presentation quality. Unlike many evaluations that focus on a single metric, AA-Briefcase tests agentic capabilities more comprehensively, exposing cases where models produce outputs that look polished but are incorrect or lack analytical rigor ➤ Built by industry experts: AA-Briefcase scenarios mirror real-world knowledge work, with tasks developed over months by experts across data science, product management and corporate strategy from companies including Google, McKinsey & Company and BCG. Task challenges are drawn from professional experience, making AA-Briefcase more reflective of the ambiguity, messy context and competing priorities that define real-world knowledge work Key results: ➤ Claude Fable 5 leads AA-Briefcase at 1587 Elo: This is followed by Claude Opus 4.8 (1356) with the next-best non-Anthropic model, GLM-5.2 (max), ~90 points back at 1266. Note that Claude Fable 5 did not use the Opus 4.8 fallback for any task in AA-Briefcase ➤ Cost per task varies by ~800x across models tested: Claude Fable 5 leads the benchmark but costs more than $31 per task on average, compared to ~$0.04 for DeepSeek V4 Flash (max). The strongest price/performance options are open weights models such as GLM-5.2 (max) and DeepSeek V4 Pro (max), with GLM-5.2 (max) scoring only ~90 Elo below Claude Opus 4.8 (max) for less than 25% of the cost ➤ Real-world complexity remains difficult for models: The top performer, Claude Fable 5, satisfies all rubric criteria on just 3% of AA-Briefcase tasks. On 31 of 91 tasks, no model scores above 50% on the rubric criteria ➤ Task difficulty scales with the number of required input files: For each rubric check, we identify the set of source files needed to pass. Across all models, pass rates fall as this file count increases, though top-tier models degrade less than weaker models More details below in thread ⬇️

Pull requests are easier to open than ever, but every review still takes human effort. Introducing pull request limits: maintainers can cap how many open PRs contributors without write access can have and set a bypass list for trusted contributors. More signal, less queue noise.

Learn more 👇 https://t.co/XrFU9MfNun

@valuetainment https://t.co/UPPUO77Ado

@swyx @midjourney @bryan_johnson @DavidSHolz @iScienceLuvr I captured a photo of you asking a question. What a front row! https://t.co/ity8V0VV0v

Can a VLM see without a vision encoder? We trained one for $100, inspired by Gemma 4 12B. Latency on an M3 Pro MacBook: 112 ms -> 1.1 ms for the image path 30% lower end-to-end image+LLM The architecture is just: patchify the image -> linear projection with pos embeddings -> LLM Writeup: https://t.co/yt0IKzsF7O