Aidan Pak @AidanPak
Premji Invest | prev @SummitPartners | prev @InvictiSecurity | @VanderbiltU math/cs alum aidanpak.com San Francisco, CA Joined April 2012-
Tweets162
-
Followers253
-
Following426
-
Likes1K
@evisdrenova This is the case for intelligent routing between closed and oss models. @OpenRouter is great and I also think @databricks has a strong case to own routing for large enterprise
It's fascinating to see how frontier capabilities seem to be first unlocked via the harness, but over time, that behavior gets inevitably baked into the weights of the model and the scaffolding around it evolves. For example, in the early days, agent frameworks wrapped LLMs in pre-defined DAGs with dedicated planning, context management, and tool-calling steps to overcome limitations in the base model such as hallucinating tool arguments, forgetting instructions, and context anxiety. These encoded assumptions quickly became stale as reasoning models were introduced and quickly improved, but importantly, that didn't mean the infrastructure around the model went away. What's super interesting about SPIRAL is that it's research on how to encode the current SOTA behavior of a harness (e.g. Claude Code dynamic workflows / loops) directly into the weights. Specifically, training a model to spin up parallel reasoning traces and then synthesize them into a final answer (which is one of the multi-agent architectures that is prevalent with dynamic workflows, e.g. fan-out-and-synthesize) but instilled end to end via RL rather than in the harness. I have been incredibly impressed by the capabilities unlocked by dynamics workflows. This is where Claude writes the orchestration as a JS program and then runs that code. The script holds the loop, the branches, and the intermediate results, fans out a pile of subagents in parallel, and a synthesizer folds them back into one answer. Other multi-agent patterns like tournaments, adversarial verification, and generate-and-filter are also prevalent. This behavior is currently a part of the harness but might soon be embedded in the weights.
The most capable reasoning systems in AI scale inference compute along several axes: sequential compute to think longer, parallel compute to sample many independent attempts, and aggregative compute to synthesize prior traces into a new improved one. But during training, we only
Great post! I think the next step for "code review" looks less like “review” and more like code validation / dynamic verification. It will be less statically reading a diff / running tests, and more simulation where agents drive an environment, run a feature against its real dependencies, check how it behaves against other services, and adversarially try to break it. The problem today is that productizing this is incredibly hard: (1) dev environments are custom to each enterprise, and (2) the cost of compute to run them at scale is still too high. I believe @cognition is heading in this direction
@JakeEpstein1 Super cool stuff Jake! Offloading as much work as possible to small, local OSS models will be increasingly important. Would be great to add in a smart router / BYO keys to route between frontier / local models
@leylndd @GeoffreyHuntley VCs when Sequoia leads the A
Longer-running agents are creating enormous KV cache that has to be moved around: offloaded across storage tiers, shared between nodes, etc. The reason CPU demand is exploding isn't that CPUs are replacing GPUs (Matmul still belongs on the GPU) which is purpose-built for it. It's that CPUs are increasingly the ones moving all that KV and data around to keep the GPUs fed along with the actual computing required to execute tool calls, navigate a browser, etc.. It's why NVIDIA's NVL72 rack pairs the GPUs with Grace CPUs.
@Jason I've heard from a source close to a Mag 7 CEO that the government is quietly pressuring large companies (Microsoft, Meta, etc.) to avoid mass layoffs. Many leaders obviously see an incredible opportunity to optimize their headcount given AI but being forced to slow it down
The utilization number you're referring to here is GPU MFU (model FLOPs utilization), which isn't a measure of whether the GPU has actually been rented out. There are no "dark" GPUs. You can see that in the price of renting the same chip (especially older generations like the A100) continuing to climb year after year. What MFU measures is how much of a GPU's theoretical compute the workload actually uses at runtime, and the answer is in that 11-45% range. Think of it as essentially the percent of time the GPU cores are doing real math versus sitting idle waiting for work, and in today's AI workloads, the core sit idle waiting on data to be moved between chips (networking bottleneck). This is especially brutal in training, where the whole job runs at the speed of the slowest GPU in the cluster. Every iteration in training requires whats called an "all-reduce," where the entire fleet has to exchange information before the next step can begin, so one straggler stalls every other chip until it catches up. A cluster pinned at 30% MFU can still be 100% leased.
@signulll I've also heard from a source close to a Mag 7 CEO that the government is quietly pressuring large vendors selling into federal (Microsoft, Meta, etc.) to avoid mass layoffs. Many leaders obviously see an incredible opportunity to optimize their headcount given AI.
Anthropic added more ARR in the past month than Snowflake, Palantir, and Databricks have added combined in their entire existence (three of the most influential companies of the SaaS era). Anthropic has roughly 1/5th the headcount of Databricks alone...
Agree that open models can probably subsume 80-90% of tasks over time (especially now with post-trained OSS >= closed-source models for a specialized task). However, I still believe 80-90% of total spend will go to the frontier as the economic value will continue to be wildly skewed toward the model that solves the hardest problems.
@djfarrelly Great post! The harness should ultimately sit outside the sandbox
Cyrus Patel @patel_cyru11219
1 Followers 11 Following
Andy Binkley @jandybin
4 Followers 85 Following
Little Emma 🌷 @LChheda61080
4 Followers 308 Following too soft for this timeline but still glowing 💗 100% follow back
knn @vc_knn
19 Followers 981 Following
sntckbj @sntckbj
40 Followers 1K Following
TheTanamiTrack @TanamiRoad
356 Followers 348 Following Macro, micro, vol and anything in between. Open minded sceptic.
alex @alexch3n
950 Followers 391 Following 20, eng @trymelius prev @google, radiant @valorant, top 500 @fortnite
crystal @crystalxtang
129 Followers 498 Following investing @ mongoDB Ventures | writing https://t.co/j4AZe5ULuM prev. passwords
Peptide Capitalist @peptidecap
108 Followers 203 Following Crossover VC (AI, sw, frontier tech) and public investor. Deploying capital into startups and gray market peptides into my body
Sofia Montoya @sofia_montoyac
83 Followers 225 Following Leading LangSmith for Startups @LangChain; prev. pre-seed through Series A investor @Crosslinkcap & growth stage investor @ Adams Street Partners
Helena Sárközi @Sarkozi24893H
13 Followers 640 Following NEODVISNI BITCOIN RUDAR 💹💹💲 TRGOVEC 📈🙏🎓 7-MESTNI ZASLUŽEK | INVESTITOR 💲🛐 DM ZA COACHING 💹💲
Col. Cinco Hamilton @cincohamilton1
117 Followers 6K Following Col. Cinco Hamilton USAF Fighter Test Pilot Former Commander, 9th Ops Group Committed to the mission
Avi Dhaliwal @dhaliwalavis
167 Followers 411 Following working on something new. previously built edtec,h, and drones and AV for the military. software guy with too much hardware experience
Cristian Raygoza @crisraygoza_
395 Followers 1K Following
sudarshan @ItzSuds
18K Followers 5K Following GP @ https://t.co/bsKudCoy24 | AI Sales @withpipeline, tried stopping wildfires @fionco
lincoln.base.eth @MurLincoln
713 Followers 768 Following Only private elite here I will mainly talk about the crypto market price action analysis etc. FOR FEW PEOPLE MAIN ACCOUNT @MurrLincoln $AMD $PLTR $HIMS $META
Jake Epstein @JakeEpstein1
18 Followers 40 Following Early stage investor @premjiinvest | Former Uber PM | Dartmouth CS & Eng | Musings on products & stuff
Andrew Wooten @andrewwooten96
186 Followers 50 Following Co-Founder & Chief Product Officer @ Rhoda AI
Robert Scoble @Scobleizer
589K Followers 52K Following San Francisco/Silicon Valley AI | Robots, holodecks, BCIs, analysis of new things | Ex-Microsoft, Rackspace, Fast Company | Wrote eight books about the future.
Active Elon CHAT @Activeonlin
154 Followers 7K Following Official Account To Privately Communicate With Elon Musk
Siddharth @siddharthvader_
1K Followers 2K Following member of capital allocation staff @scalevp | previously ML, product, and engineering | math & cs @columbia | I ♥️ SF & bay area sports
Michael LaFramboise �... @LaFrogman
4K Followers 603 Following Founder @ https://t.co/PaJaxMS5KA - 🙏🏋️👨💻 into cows
x402 Intern @x402intern
2K Followers 1K Following x402 Intern @x402 @x402foundation @Solana @Base @AIBTCdev @Celestia. x402intern on TG. DM for all things x402
Madeline Chen @Madeline_Chen_
90 Followers 474 Following Life is a journey I enjoy history yoga fitness saunas and red wine 🍷 or sparkling wine 🍾 We can talk about books or discuss the best retirement destinations🏖
PoEWit Technologies I... @PoEWitTech
74 Followers 73 Following PoE lighting + security infrastructure. One cable. Cloud-controlled. Real-time response.
Shubhan Dua @defi_dua
7K Followers 5K Following Growth @quizlet | Built & exited @answersai_ai (2M users) | Prev @artemis @ucberkeley 24 @dormroomfund Ilahi
Carrie Underwood Acce... @carebarer
108 Followers 1K Following Isaiah’s & Jacob’s mom. Also, mom to fur babies Penny Jean, Zero, & Charlie. Blessed & grateful!
Alan Zhao @zhaolan_monk
109 Followers 620 Following Cofounder & Head of Product & Marketing @ Warmly | YC | GTM Context Agents
Based Thought @based_thought
121 Followers 437 Following ML focused software engineer. AI infrastructure investor
sabina smith @Sabina_Smith_
470 Followers 518 Following Thinking about Something and what that Something means for Something Else
Ben @benftingle
0 Followers 143 Following
Sara @37Ej6h53qL0Tq9b
37 Followers 1K Following Don’t let anyone dim your light simply because it’s shining in their eyes.
Vedhika Sharma @Vedhika_143
17 Followers 1K Following
Kebba @Kebba197634
194 Followers 7K Following Jesus Christ say: I am the way and the truth and the life.十十
Databricks @databricks
93K Followers 1K Following Databricks is the Data and AI company, helping organizations build and scale data and AI apps, analytics and agents.
Michael Trinh @_mctrinh
9K Followers 73 Following researcher @UUtah, founder @mindlinkor, https://t.co/DeNqbAi9an 🇻🇳🇰🇷🇭🇰🇮🇹🇯🇵🇺🇸
Cyrus Patel @patel_cyru11219
1 Followers 11 Following
Maya Agnihotri @mayarox5
469 Followers 71 Following growth & partnerships @trymelius | prev swe @nike | ur fav hot girl tech bro
Yuchen Jin @Yuchenj_UW
192K Followers 793 Following Cooking fun AI systems & products @databricks. Prev: co-founder & CTO @ Hyperbolic, OctoAI (acquired by @nvidia) Apache TVM, PhD @ University of Washington.
Jubayer Ibn Hamid @jubayer_hamid
1K Followers 197 Following PhD at Stanford CS, SR at Google. Previously: maths/physics at Stanford.
cat @_catwu
93K Followers 392 Following claude code + cowork @anthropicai, prev: @dagster, @scale_ai
Anjney Midha @AnjneyMidha
48K Followers 3K Following founder @amppublic • visiting scientist @stanford • teaching @cs153systems
Jarred Sumner @jarredsumner
178K Followers 644 Following building @bunjavascript at @anthropicai. formerly: @stripe (twice) @thielfellowship. high school dropout. npm i -g bun
unusual_whales @unusual_whales
4.4M Followers 2K Following Stocks/Options/Crypto/Market News/Tools. Not advice @Polymarket partner Open a tastytrade account: https://t.co/wGf2ZdlXpw Discord: https://t.co/0xJ9e0ZYYG More: https://t.co/nsxZlPV0pC
Alberto @taiuti
5K Followers 946 Following CEO & co-founder @reactorworld - super hardcore multi theft auto san andreas player
Aditya Jayaprakash @adijayaprakash
579 Followers 1K Following CEO / Professional CI therapist at @useblacksmith
Noam Brown @polynoamial
146K Followers 922 Following Researching reasoning @OpenAI | Co-created Libratus/Pluribus superhuman poker AIs, CICERO Diplomacy AI, and OpenAI o-series 🍓 reasoning models
Evan Feinberg @enfeinberg
1K Followers 170 Following Founder & CEO, https://t.co/JvqkTU20qB / Scientist developing deep learning methods for drug discovery / Pande Lab @ Stanford. Applied Physics @ Yale.
TK Kong @tkkong
12K Followers 595 Following Building something new. Previously @RampLabs, founder/CEO @VenueHQ (backed by Sequoia, acq by Ramp). Tennis fan. 🇰🇷
Byron Deeter @bdeeter
22K Followers 588 Following Partner @bessemervp w/ @servicetitan @anthropicAI @perplexity_ai @twilio @intercom @waymo @procoretech @clickhousedb @canva @maintainx @49ers
alex @alexch3n
950 Followers 391 Following 20, eng @trymelius prev @google, radiant @valorant, top 500 @fortnite
crystal @crystalxtang
129 Followers 498 Following investing @ mongoDB Ventures | writing https://t.co/j4AZe5ULuM prev. passwords
Pocket Jacks Capital @rl_env
2K Followers 220 Following rick rubin of in the flow members of technical staff
Peptide Capitalist @peptidecap
108 Followers 203 Following Crossover VC (AI, sw, frontier tech) and public investor. Deploying capital into startups and gray market peptides into my body
seb (recruiter arc) @hiiinternet
8K Followers 2K Following Head of Talent Eng @fonziai, prev: cookd cofounder (acq), @whop 1st recruiter. recruiting engineers for startups see website 👇
Ramtin Naimi @ramtinnaimi
7K Followers 689 Following Founder & General Partner at Abstract @abstractvc
Evis Drenova @evisdrenova
5K Followers 929 Following Principal Eng @ Entire | prev. co-founder @ Neosync (acq. Grow Therapy) @ycombinator @uchicago
Irrational Analysis @insane_analyst
42K Followers 101 Following Engineering-Driven Investment Analysis
Dane Knecht 🦭 @dok2001
40K Followers 2K Following I help invent the future. CTO @cloudflare. Angel investor.
Invicti Security @InvictiSecurity
3K Followers 232 Following We deliver the only proof-based application security platform that finds, validates, and prioritizes real vulnerabilities before attackers can exploit them.
Meryem Arik @MeryemArik9
739 Followers 811 Following Inference Lab. I’m the token (generation) woman - CEO/Co-founder Doubleword - F30U30 (anti-signal?)
Sofia Montoya @sofia_montoyac
83 Followers 225 Following Leading LangSmith for Startups @LangChain; prev. pre-seed through Series A investor @Crosslinkcap & growth stage investor @ Adams Street Partners
Oliver @olvrgln
3K Followers 1K Following when I was a boy I heard God’s voice. now I build b2b saas @ https://t.co/us8mCfGrFf
Ben Warren @bwarrn
2K Followers 1K Following founder @mesa_dot_dev, future governor of Moon. prev: yc, census, microsoft, northwestern
sudarshan @ItzSuds
18K Followers 5K Following GP @ https://t.co/bsKudCoy24 | AI Sales @withpipeline, tried stopping wildfires @fionco
Kyle Jeong @kylejeong
7K Followers 541 Following growth engineering @browserbase | also @vestucla @kp_fellows math/cs @ucla
Aparna Dhinakaran @aparnadhinak
10K Followers 2K Following founder @arizeai. I post about agents and evals. previously: @uber, @ycombinator alum
Modal @modal
32K Followers 160 Following AI infrastructure that developers love 💚 Run inference, sandboxes, batch processing, training, and many other things on Modal
kate @whoiskatrin
23K Followers 909 Following building agents and sandboxes @cloudflare, so you don't have to
Melius @trymelius
1K Followers 20 Following The agentic harness for creatives Be the tastemaker as your agent builds your generative workflows.
Sam Ragsdale @samrags_
9K Followers 1K Following CEO @merit_systems -- Open Agentic Commerce Products: @poncho_ai @x402scan @agentcashdev Retired zkVM guy @a16z
Ethan @ethanruhe
274 Followers 680 Following agents mostly | ai product lead @temporalio | dm with product feedback
Viv @Vtrivedy10
13K Followers 2K Following applied research @LangChain Labs, prev @awscloud, phd cs @templeuniv






















