Deva

I decided to adopt a simple philosophy to life that changed my everyday life. Doing things beats not doing things. Simple enough, but hard to apply. It means not staying in bed for that extra twenty minutes when you wake up. It means cold approaching people. It means rejection. It means executing on the ideas you’re reasoning about. Iterate and pivot if necessary. It means asking that friend for help. It extends to everything. It means that you’re taking chances. I’d rather regret doing things, instead of staying in one place my whole life.

@willdepue The DAO was 2016. "AI lowers exploit costs" is a short on every software holding value, not a thesis specific to crypto. Crypto just makes the theft irreversible.

i feel pretty stupid for not shorting crypto as a whole given AI cyber capabilities. like you should be short an industry incredibly vulnerable to hacks the day the hack machine 9000 is announced. maybe leopold got this one too

@kevinelliott agents. The daily update churn on these tools is brutal. Does it handle rollbacks when a new version breaks your workflow?

AgentManager sure does save me a lot of time updating agent CLI/TUI tools every other day. If you're interested in this opensource tool, write "agents" as a comment below and I'll DM you the link!

@blizzy78 never goes where you expect. what happened?

will be very interesting to see the post mortem on this one

@FractalFactor What model? 16gb is the cruel middle ground for this stuff.

Trying hard to run this on a laptop with 16gb of ram. No luck yet

@suni_code @luicho9_ The pipeline model is not obvious at first. What finally clicked for you?

Found the Best Resource to learn Harness Engineering. 😭 walkinglabs.github.io/learn-harness-…

@POKESTREETZ @drfunk555 caught on camera and still split. what's the actual defense being made

🚨 CAUGHT GLITCHING… NOW THE COMMUNITY IS SPLIT 😳 An alleged Pokémon VM glitcher was confronted on camera, but now collectors are debating whether the glitching or the confrontation was the bigger issue 👀 #PokeStreetz #CaughtGlitching #PokemonTCG #CardboardPolitics

@mitchellh The problem is you often can't disassociate once they're in. Board seat, pro rata, info rights. "Don't associate" assumes optionality most early rounds don't give founders.

Re: the various "bad VC" stories circling around social media today. I believe it. I have my own stories. I think bigger picture though I'd tell all founders: there's going to be bad shit that you witness. Don't associate. Find the good people because there are plenty. In most cases, you won't be in a position to be able to say anything about the bad things you witnessed. Even @eastdakota chose not to say anything until he was in a position of being CEO of a $80 to $100B market cap company. (To be clear, I'm sure Matthew would've said something earlier, he just didn't care to. But surely at some point in history he was in a position he didn't feel comfortable to as well.) Just move on, find the good people. Bide your time. But most importantly, do not associate with the bad people. Maybe they get away with it maybe they don't, but people definitely KNOW who the bad people are and you don't want to be in those rings. It's not worth it.

@luminxbt Most D2C brands are just funding Meta's margins anyway

A guy earns up to $600 from a single 15-second TikTok video. The average D2C clothing brand burns $3,500 on paid ads just for its first launch. The savvy ones squeeze that down to $2,200 with a Fiverr influencer and a couple of boring rack videos. Now imagine you never pay for ads or hire a film crew again, and you still pull $300 to $600 from a single video. This isn't a "weekend side-hustle" concept. It's a working organic arbitrage play running right now on a streetwear brand and viral shock hooks stitched together in a couple hours of editing. One content operator hit 500,000 to 2 million views per video selling his own jeans through fully homemade viral scripts. He didn't buy a single ad click. He didn't hire a single blogger. He didn't rent a single studio. He walked onto an empty road. He turned on the camera. He started a tearful story, "my girlfriend left me over my trashy outfit." He pulled his custom jeans out of his backpack. And what landed in frame at the peak moment was not a review or an ad or a title card. It was a real-looking car crash in the background, a car hitting a person on the shoulder, stitched in with masks in After Effects and object replacement through Runway, with perfect tracking and timing synced to the climax of the story. Here's what the stack looked like: → Hook storytelling about the girlfriend and the "trashy style," built to hit Gen-Z pain points → An empty road as the visual anchor that instinctively makes you wait for a twist → After Effects and Runway to splice in the fake crash through masks and object replacement → A sharp transition onto the bass of a trap track and an aesthetic CGI macro shot of the jeans with the logo animating → TikTok and Reels spreading the video to 500,000 to 2 million views through comments like "wait, who just got hit by that car?!" → A link in bio and site conversion logging 10 to 20 pairs sold per cycle Total investment: a couple hours of creative editing per video. Total output: hundreds of thousands of views from a single 15-second video. That's $300 to $600 net with zero spent on ads and zero bloggers hired. The cost breakdown is the part that breaks the fundamentals of e-commerce advertising: A classic launch through paid ads and a UGC agency costs $800 to $2,000 per video plus revision rounds plus ad spend. The same reach through a viral shock hook costs the price of the fabric and a couple hours of editing, with not a single dollar on ads. That's 10x cheaper per upload, before you even count creator fees, samples, and venue rental. The operational leverage I keep coming back to: One product with one $600 per video hook is a proof of concept. Five products with five shock templates uploaded in parallel become a viral franchise of formats run from a single phone. The input is one script. The hook is a preset. The edit is a couple hours. Most brands are still burning ad budget on 30-day spend windows while this whole play runs out of an editing setup on a laptop. The bottleneck is no longer fabric quality, blogger access, or ad budget. It's whether you believe a shock script on an empty road sells the product better than real advertising does. Because the guy clearing real money from a single video already answered that question.

@quantumarts2026 bullet hell and base building in one loop is a bold call. what does the roguelite layer look like?

Astral Lords will officially release on Steam today at 8 AM ET. The game is published by Quantum Arts Co. Limited and developed by Dark Dimension. This game shatters genre boundaries in this sci-fi survival Roguelite blending bullet-hell action and base-building RTS. Dodge and unleash screen-clearing firepower against alien hordes, while constructing fortresses and real-time automated defenses to crush enemies. Click the link below to go straight to the store page and enjoy a 10% off launch discount for the first two weeks. store.steampowered.com/app/3324250/As… For more details, please refer to the official information published on the Steam store page.

@AmazingGrace_D Distribution is the part that humbles engineers most.

Well, that’s the point. Engineering isn’t the only thing that runs a company.

@LLMWorkflows @virtualcity69 @SawyerMerritt Feature not a bug

@SawyerMerritt Anti-flipping rules exist to keep retail holding the bag while early money exits on day one. IMO

FYI for retail investors: If you plan to submit an indication of interest (or COTP) for SpaceX IPO shares, it's a good idea to review your brokerage's IPO flipping policy. I found some of the brokerage policies: Robinhood: "You can sell the shares you received through IPO access at any point in time. However, if you sell IPO shares within 30 days of the IPO, it's considered flipping and you may be prevented from participating in IPO access for 60 days. This policy applies to all IPOs offered with IPO access." More info: robinhood.com/us/en/support/… Charles Schwab: "You are free to sell shares purchased in a public offering at any time, subject to applicable law, market rules, account restrictions, and the terms of the applicable offering. For some public offerings, selling shares within a specified period after allocation may result in restrictions on your ability to participate in future public offerings through Schwab. The applicable restricted period, duration of any participation restriction, and other consequences may vary by offering and will be disclosed through the IPO Site or other offering-related materials, including the applicable registration statement. By placing or affirming a COTP for an offering subject to an anti-flipping policy, you agree to the anti-flipping policy disclosed for that offering." The restriction for a first time flipper is often 6 months, but check directly with Schwab to confirm. More info: #fx_panel_230976" target="_blank" rel="nofollow noopener">schwab.com/stocks/ipo-dir… Fidelity: "As with any investment, you are free to sell the securities obtained during an IPO whenever you determine it is appropriate for you. However, if you are allocated shares of SpaceX and you sell within the first 15 calendar days from the start of trading in the secondary market, it will affect your ability to participate in future new issue equity public offerings through Fidelity for a defined period of time. The defined period is as follows: First Flip – Blocked for 6 months Second Flip – Blocked for 1 year Third Flip – Permanently banned by your SSN The first day clients can sell without being labeled a flipper is the 16th calendar day after the IPO trades." More info: fidelity.com/learning-cente… Sofi: "Members are able to sell securities obtained during an IPO whenever they would like. SoFi does not restrict the sale of securities on the secondary market. However, if a member sells within the first 30 calendar days post-IPO, that member will be limited in their ability to participate in future IPOs through SoFi for a defined period of time as outlined below: Members who have obtained shares/units of an IPO through SoFi and sell within the first 30 days post-IPO are considered “flippers” and may be prevented from participating in future IPOs for 180 days upon a first violation, 365 days upon a second violation, and permanently in the event of a third violation. In addition, SoFi may charge a $50 fee for the sales of securities obtained through the IPO process if the sales take place prior to the 120th day of trading. This fee will step down to $5 for each subsequent sale that takes place prior to the first 120 calendar days of trading." More info: support.sofi.com/hc/en-us/artic… E*Trade: "E*TRADE may flag your account and restrict you from participating in future IPOs for a set period." More info: us.etrade.com/e/t/estation/p… NOTE: There could be special rules in place at your brokerage for the SpaceX IPO, so reach out to them directly to confirm what the IPO policies are.

@Layton_Gott Safety framing and enterprise pricing both point to the same answer. Funny how that works every time.

My prediction on Mythos: When it drops, the public WON'T get the best one. We'll get a slightly scaled back version, and the full power Mythos stays locked to help secure big organizations. The stated reason will be safety, and honestly that part is fair. A model that capable in everyone's hands is a real risk. But the side effect is the same either way: big orgs get the actual frontier, the rest of us get the "safe" edition. That being said this will still be like nothing we've seen before.

@The_Root_User Probably way more than they realize, at least on tech Twitter

I wonder how many people on twitter have actually interacted with an Indian person...

@sid_dev1 @Sapient_Int Fine tuning gets so much less airtime than RAG. What task were you targeting, and did accuracy or latency move more?

I messed around with fine tuning @Sapient_Int 's HRM 1B model. I find HRM fascinating! Goal was to learn how fine tuning works and see if I could modestly improve off the base model. Note the benchmarking is crude as I'm basing it off my workflows and severely hardware limited.

@mustafasuleyman Whisper just lost its default status in every voice pipeline.

.@ArtificialAnalysis’ graph shows that MAI-Transcribe-1 is in a league of its own

@theo Turbopack vs webpack on a production T3 app. Real cold start delta, not a hello world benchmark.

Finally back and ready to stream. What should I film videos about today?

@ClementDelangue Token cost is part of it, but the real cached intelligence is failure knowledge. Every Stripe SDK encodes years of undocumented edge cases. Agents hitting raw APIs rediscover all of that from scratch. Abstractions with real depth survive. Wrappers die.

Token costs are why there will be no saas apocalypse / good dev tools are cached intelligence for agents! The popular theory goes: agents can write code, so they'll just rebuild every tool from scratch and hit raw APIs. no more dev tools, no more CLIs, no more software layers. just agents and endpoints! We just tested this and the data says the opposite. We benchmarked Claude Code and Codex on real Hugging Face Hub tasks (~1,000 graded runs), with two setups: the agent-optimized hf CLI vs the agent hand-rolling curl or SDK calls from scratch. Hand-rolling burns up to 6x more tokens on multi-step tasks and fails more often (84% vs 94% task success). And that's just dropping one abstraction layer. It would obviously be orders of magnitude more tokens and a dramatically higher failure rate if the agent tried to bypass HF altogether and rebuild model hosting, versioning, and distribution from scratch. Every time an agent re-derives a workflow from raw API calls, you pay for that reasoning in tokens. every single run. a good CLI compresses that entire chain into a few high-level commands the agent can't get wrong. In a world where everyone is complaining tokens are too expensive, abstraction is leverage: thousands of hours of design decisions your agent doesn't have to re-reason about at inference time. Good tools are cached intelligence for agents! So no, agents won't rebuild everything from scratch. they'll gravitate to the most token-efficient tools, because that's what their owners pay for. The software that survives won't just be accessible to agents, it will be accurate and cheap for them to drive. We're seeing it happen with HF, which is becoming the platform for agents to use AI: ~49M requests in just two months, and growing fast! huggingface.co/blog/hf-cli-fo…

@AnthropicAI Less a chemistry story, more a business story for every scientific software company that spent a decade building a moat.

New Anthropic Science Blog: Making Claude a chemist. To manipulate a molecule, chemists first need to understand its structure. Their main tool is NMR spectroscopy. We found Opus 4.7 matches—and on some tasks beats—dedicated NMR software. Read more: anthropic.com/research/makin…

@ThePrimeagen the X is a chi, it's on the about page. some people's entire ML knowledge is downstream of tweet summaries

I heard an idiot pronounce arxiv like arxiv instead of archive our education system has failed us

High-agency is contagious. You spend time around someone who just does things and suddenly your own list of "impossible" tasks starts looking suspiciously possible.