AI for the physical economy

San Francisco, CA
.@TryPallet's confidence was restored during its hardest stretch. @SushanthRaman and BCV Partner @kevinzhang on what it took to get there, and why the desire to survive can take you far. piped.video/sM_L9zC1UYw?si=qEMD…
1
7
728
Hey @JudgmentLabs come get your lil guy
Made with AI
1
5
177
Thanks for the 6 free resets @sama
1
1
24
1,392
Pallet retweeted
Im in Jev heaven (Jevean?) At Pallet we're building automations for many of the world's largest enterprise logistics companies. On a customer classification experiment, I compared Jev Only vs Hybrid Jev LLM vs prod LLM data The results are INSANE
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev • 20-200x faster • 40-400x cheaper (w/ output tokens free) • Frontier composable intelligence optimized for decisions AFAICT the shortest path to AI-based economic revolution
1
2
7
247
Shipped Custom Models together. Full-stack partnership means inference and something to wear to standup. Ty @baseten 🫶
1
3
257
When I was job searching six years ago, I used this list to figure out where I wanted to work. It comes full circle.
Replying to @chrisbarber
Breakout List, all picks (2 of 4): 51-100 people: - Aaru (@seekingtau, @virtualned, @john_k) - Beacon Software (@nilamg, @divyahansg) - Fab2 (@szeloof, @jimkxa) - fal (@burkaygur, @gorkem) - Listen Labs (@itsalfredw, @florian_jue) - Medra (@michellearning) - Pallet (@SushanthRaman, Andrew Spencer) - Periodic Labs (@LiamFedus, @ekindogus) - SF Compute (@evanjconrad, Eric Park, @Ethan_is_online, @EricMenees, @apagajewski) - Simile (@joon_s_pk, @msbernst, @percyliang, @ElainaYallen) - Twenty (Joe Lin, Leo Olson, @skyleronken, Pete Sorrentino) - XDOF (@philippswu, @YideShentu, @itsnemojin) 101-250 people: - Antares Nuclear (@jordanbramble, @juliadewahl) - Aven (@SadiSKhan, @murtada, @collinwikman, @AlmazNanjappa, Jeremy Solomon) - Bitwise (@HHorsley, @hongkim__) - Boom Supersonic (@bscholl, @jkrall, @joe_wilding) - Brain Co. (Dan Ashton, @ericwu01, @eladgil, @jaredkushner) - Braintrust (@ankrgyl) - Castelion (@hargsb, @sean_pitt_, Andrew Kreitz) - Enter (@mateus_cr123, Henrique Vaz, Mike Mac-Vicar) - Exa (@WilliamBryk, @jeffzwang) - Factory (@matanSF, @EnoReyes)
1
2
20
8,071
HITL getting out of hand
1
4
213
We're honored to be part of this list 🫡
I made a list of great startups to join. It's called the Breakout List. The list has 92 companies. These are the 20 with 25 or fewer employees: - Hone (@moritz_stephan, @CarloWillem, @oqbrady) - Normal (@ansonyuu, @hudzah) - Standard Intelligence (@G413N, @devanshpandey) - Tacit Labs (@ninklefitz, @AmDroste) - American Terawatt (@atroyn, @rslparker, @aranibatta) - Conduit (@clemvonstengel, @riopopper) - Convergent (Omkar Savant, Vivek Katara, @debnilsur) - Core Automation (@MillionInt, @_arohan_) - Engram (@dan_biderman, @EyubogluSabri, @realJessyLin) - Instinct (@noahrshinn) - Keenable (@styskin, Matthias Petri) - Lumaril (Mark Elliot, Ben Duffield) - Neion Bio (@Dimkell, Sam Levin) - Pangram Labs (@max_spero_, @bradley_emi) - Quadrillion (@echinaceous) - Re (@karnsaroya, @AnandDhillon, @thecliffwhite, @benaneesh) - Ricursive (@annadgoldie, @Azaliamirh) - Sail Research (@neilmovva, @blintzbase) - Trajectory (@rronak_, @michaelelabd, @QuantumArjun) - Watney Robotics (Sean Cheong, Ryan Gannon) Picks from Elad Gil, Charlie Songhurst, Keith Rabois, Mike Vernal, Alana Goyal, Sonya Huang, Ramtin Naimi, Marc Bhargava, Cory Levy, Aashay Sanghvi, Konstantine Buhler, John Luttig, Varun Gupta, Ray Tonsing and Avichal Garg. Disclosure: I'm a small investor in American Terawatt, Convergent, Standard Intelligence and Trajectory (in this post), and in Factory, Physical Intelligence and SF Compute (elsewhere on the list). I didn't vote. The full list is on Breakout List.
1
6
332
Sometimes building an outlier company means walking away from a good one. The first version of @TryPallet was working. But Sushanth Raman believed it needed to change to deliver truly ambitious impact for the global supply chain. He pushed ahead with a pivot despite resistance from customers, teammates, and investors. @kevinzhang and @SushanthRaman discuss what drove the decision, what it cost, and what came next. piped.video/sM_L9zC1UYw?si=dWh0…
4
38
25,713
Zero tokens were used in this build
1
11
1,245
We're proud to partner with @onelineage, the world's largest temperature-controlled logistics company. Today, @trypallet AI agents process two million of their shipments a year at 99% accuracy, helping them scale without compromising the high-touch service they're known for.
2
1
7
891
Our team in field notes
很喜欢这种把旅行照片收进田野笔记里的方式。 照片留住现场,旁边那枚略微缺墨的彩色橡皮章,只刻下最值得记住的轮廓。 树、穹顶、海岸都没有被讲满,反而像旅途中随手盖下的一页,安静,又很有地方感。 prompt: 请将我上传的每一张照片分别制作成一张独立的「橡皮章旅行田野笔记海报」,每张照片单独输出,不要多图拼贴。 整体采用4:3横版构图,将画面划分为左右两个区域,但不要绘制明显的分隔线。 左侧约占画面58%,忠实保留原始照片。准确保持主体身份、地形、建筑、植物、人物、空间关系、自然光影、真实质感和原有色彩氛围,仅进行克制的艺术出版物级摄影调色,并加入极轻微的细腻胶片颗粒。为适配版式可以自然裁切,但不得拉伸、扭曲、移动、替换或重新绘制主体。 右侧约占画面42%,使用温暖的米白色旧纸作为背景。纸张具有细微纤维、天然颗粒、轻微使用痕迹和哑光触感,同时保留大面积未经印刷的纸张留白,让空白成为版式的重要组成部分。 分析原始照片,从中提取最具地点辨识度的主体轮廓、建筑结构、地形走势、植物姿态、道路、水岸线或其他关键视觉关系,将其压缩成一枚小型多色橡皮章图像。 不要逐项复制照片中的全部内容。只保留能够让人一眼认出原始地点、主体和场景关系所必需的最少信息。删除人群、车辆、密集窗户、重复建筑、细碎植被、装饰构件和无关背景。 章印位于右侧纸张区域的中下部,整体只占右侧区域高度的约30%—38%,周围必须保留充足留白。章印不能放大成普通插画、完整风景画或品牌Logo。 根据原图构图决定章印的组织方式: - 标志性建筑:保留最有辨识度的外轮廓、屋顶、穹顶、拱门、塔楼或主要结构。 - 山地聚落:将建筑压缩成沿地形排列的少量阶梯状色块。 - 海岸风景:保留山体走势、聚落层级、岸线和少量断续水纹。 - 城市远景:保留主要天际线、一个标志性建筑及一两层远山。 - 自然景观:保留主要山体、树木、水岸或道路的方向关系。 - 近景遮挡物:如果其对原图叙事重要,可作为前景章印轮廓保留。 从原始照片中提取2—4种专色油墨。优先使用炭黑、深绿、砖红、赭黄、灰蓝、灰褐等经过降低饱和度的颜色,但不得强行套用固定色盘。保留原照片最有辨识度的色彩性格,仅允许一个小面积颜色作为视觉强调。 每一种颜色都要呈现为分别手工盖印的效果: 真实橡皮章雕刻纹理、手工刻痕、粗细不均的排线、轮廓缺口、断裂边缘、干燥缺墨、纸张透底、颗粒状油墨、压力不均、局部重影,以及约1—2毫米的轻微套色偏移。 不同色层之间允许出现自然错位,边缘不能数字化平滑。印迹应像真正刻制后压在旧纸上的橡皮章,而不是套用滤镜的照片、平滑矢量插画或线稿Logo。 根据照片中的地点、主题和视觉意象生成文字: 地点英文名称 No. 编号 三个简短英文关键词 西元年份 文字放在章印下方或邻近留白处,使用小型、克制、略带机械误差的打字机字体。排版应像旅行者的田野记录,而不是广告标题。确保所有文字拼写准确,不添加无关标语、品牌或装饰性文案。 整体气质像建筑师、旅行作家或自然观察者保存的田野笔记:安静、克制、触感真实、地域明确、带有手工误差和收藏感。照片负责记录现场,章印负责留下记忆中最值得辨认的部分。 Avoid:明显的中央分隔线、圆形印章、中文红色印章、邮票齿孔、蜡封、贴纸拼贴、旅游纪念品模板、平滑矢量Logo、通用城市图标、完整复制所有建筑、密集刻画、儿童手工感、卡通风格、3D渲染、塑料质感、光滑数字渐变、过度饱和、过多文字、装饰堆积,以及重绘或改变左侧原始照片。
Made with AI
1
9
1,037
Pallet retweeted
Bless up and ship more pallets
Just got blessed by @JudgmentLabs
1
1
14
4,052
Just got blessed by @JudgmentLabs
1
1
17
5,682
Pallet is the fastest-growing AI-native company in logistics. I asked their founder what he attributes that to
Simply automating back-office tasks without directly tailoring those tools to your specific company workflows results in minimal margin improvement. Sushanth Raman explains how Pallet avoids this trap by ensuring every product asset focuses strictly on boosting user margin and network execution. Shifting your digital strategy to highly specialized AI agents that securely process institutional workflows protects long-term company returns, stabilizing primary shipping lines even as industry costs fluctuate.
6
2
32
10,566
Today, we're introducing Browser Automation for @trypallet AI agents. Supply chain teams rely on carrier portals from @Maersk, @Saia_Inc, and other transportation providers, but these websites were designed for people. Many don't expose reliable APIs, and even when they do, the last mile of execution still depends on someone clicking through the browser. Pallet agents can now securely operate carrier portals, customer websites, and appointment systems, completing browser-based workflows from start to finish.
4
1
6
482
We've seen our AI token spend skyrocket 10x over the past year as more customer workflows move into production. At some point, inference spend crossed a threshold where it no longer made sense. That's what led us to train our first production model. Here's what we learned: – The financial break-even point between training your own models and using frontier models is around $750/day. – The dense 27B model beats a MoE model with more parameters. Smaller memory footprint left more room for long contexts and KV cache. When memory is the bottleneck, fewer total parameters beat fewer active ones. – The single biggest improvement came from label taxonomy. About 18% of our production traces were assigned to a catch-all "I can't process this" label. The model learned to rely on that fallback more than it should. Removing that single label increased accuracy from 84% to 95%+.
1
1
8
2,533