Jacob (@jvboid) header
Jacob (@jvboid) profile picture

4

San Francisco, CA
Joined November 2022
Photo by Jacob (@jvboid)Photo by Jacob (@jvboid)Photo by Jacob (@jvboid)Photo by Jacob (@jvboid)Photo by Jacob (@jvboid)Photo by Jacob (@jvboid)Photo by Jacob (@jvboid)Photo by Jacob (@jvboid)Photo by Jacob (@jvboid)Photo by Jacob (@jvboid)
Jacob retweeted
发现一个特别好笑又扎心的事:现在的 AI Agent 跟职场新人一模一样——不记笔记的时候全凭本事干对,一旦开始自己写“错题本”,反而越干越废。 刚刚刷榜拿下 HuggingFace 24小时与7天双料第一的 KAIST 新论文(WikiTrace),干脆利落地扒下了 Agent 记忆系统的底裤: 那些三次裸考都能全对的简单任务,加上“历史经验”之后居然直接翻车了。团队拉出 355 个翻车现场一查,高达 52% 的弱智操作,全是因为 AI 死板照抄了自己以前留下的“半吊子经验”。 这就是离谱的“记忆致残(Memory-Induced Degradation)”。👇
16
20
2
285
17,152
Replying to @0bNARA
4
2
18
4,163
Jacob retweeted
Hard to overstate how big this is. tl;dr - CVE-2026-98432: a remote use-after-free in the Linux IPv4 TCP handshake (kernels 4.4+). Any unpatched server that accepts IPv4 TCP on an MPTCP listener can be taken over as root, with no credentials. Go servers built with Go 1.24+ use MPTCP listeners by default. Ubuntu and RHEL have no fixed kernel yet. Check with ss -Mlnp. Quick fix for Go: GODEBUG=multipathtcp=0.
35
93
16
1,335
196,073
Jacob retweeted
Some predictions: >Increasingly likely that Anthropic will postpone IPO or numbers will come substantially below target (200b on 2T) >Data markets will get hit harder than compute markets by the cool down >Both will likely IPO by end of 2028. Too large to stay private
9
10
4
89
25,655
I hope this is a wake up call for a lot of you. Tech no longer provides you the safety you once desired, so that exact fear you ran away from is something you'll face in the next couple of years. Startups provide the biggest risk/reward ratio for your career and provide the same compensation in salary. Don't expect to get the same from big tech or corporate. Go build something that changes the world.
October’s layoffs so far: Netflix: ~1,000 (6%), mostly product and tech eBay: ~1,000 (9%), its second round this year HubSpot: ~660 (7%) EA: ~400 (3%), and BioWare is closing Dropbox: ~200 (9%), plus its CTO Amazon: hundreds across seller support, engineering and customer service Airtable: a big chunk of the team, after Bending Spoons bought it Microsoft: buyout offers to ~60,000, about a quarter of the company
4
7
1
245
78,871
We brought a knife to a gunfight — one agent and quantum circuit against 40 agents swarm. I walked into the Gemini 4 Argon hackathon expecting to work on a voice and camera economy factiorio game. Teams were competing on how many agents they could keep up with. One team orchestrated 40 agents to create a playable game in 5 hours. Others cloned full games from single YouTube video, pulled hardware BOMs of research screenshots, and automated entire wet lab from a photo of a real lab bench. Times changed, hackathons had moved from clever code to high velocity token allocation. Then we sat down with Hanah Rahman, a UC Berkeley quantum researcher and @jvboid. Hanah can hold Hilbert spaces in her head yet was locked out of agent ecosystem simply because no univ teaches domain scientists what a context window costs, how tool contracts fail, or how to built an harness. So we teamed up to build My Little Brain to process 8 EEG channels into 10 band powers, mapping 4 biophysical tuning knobs (4 settings each, 2 bits per knob) onto an 8-qubit register. Picture a library of 256 small brain states benchmarked against CHB-MIT clinical seizure data. Pull books at random and you open about 128 before you find the target. Rather than brute forcing all 256 states, Gemini 4 Argon wrote QASM circuits over 5 iterative turns on IBM simulator, agent ingested top-8 rank verdicts, pruned away non-viable branches, and compressed its own circuit from 18 two-qubit gates down to 4. We wanted to see how an autonomous compiler could navigate physics-driven search spaces under strict hardware noise constraints.
3
2
9
371
Jacob retweeted
I agree. Robots are the non-biologist's abstraction of how scaling in biology scales. Imagine people in the 1940s pouring money 100x money into automating the vacuum tube factory than inventing the transistor. Imagine putting everything into giant CPU factories instead of inventing the GPU. We are disproportionally thinking/resourcing innovation in biology as if it's car manufacturing. There is ample opportunity to improve, redefine and scale measurement in front of us. I recently talked to someone at a robot-first science factory with near billion in funding boasting how they measure ~1000 mutants a week. A team of 4 at Dyno scaled the equivalent assay multiplexing by 25x. 1 person can run it. The robot is scaling the wrong layer. I'm not against robots and I think they will have the role to play. But I think their impact on biology is overblown, especially on how it scales in performance per dollar deployed. The danger is we become so obsessed with vacuum tube factories that we don't fund the transistors of biology. Complex physical value chains are somewhat path dependent, once you build a highway instead of a subway, you have to deal with a lower efficiency, higher pollution world. Most current robotic approaches are a crude, and poorly scalable, way to scale biology.
Maybe controversial: the conversation about scientific bottlenecks is overfocused on automation. New measurement technologies deserve more attention. You could automate an assay that needs an entire facility – or you could REINVENT the assay so that it fits into just a few tubes.
28
54
13
354
30,910
Feels like March 1, 2020 for Cyber. Not normal, something is happening. Internet lockdowns in a couple of months?
67
182
81
2,306
396,398
[continued] I say "almost" because there is going to be one way to maintain secrecy. Software as a service with the executable hiding behind a network connection to the cloud. But as of now, we must assume that anything shipped as an executable, or even a firmware image, is as transparent as glass. It won't keep its secrets for any longer than it takes for somebody to be motivated to throw an LLM at it. I did not see this coming. When I wrote down the theory of open source, 30 years ago now, I thought closed source would be gradually driven towards extinction by cost gradients, but survive indefinitely in certain niches. I did not foresee it being wiped out in a technoapocalypse. Ironically, I thought one of the application areas in which closed source would persist longest was games. There are still some obstacles. In US law, decompilation of a binary is a derivative work of the binary that falls under whatever copyright it had. However, it is also settled law that if you decompile binary code to a precise specification of what it does, then generate fresh code from that without looking at the decompiled stuff, you're in the clear. This was the case law that allowed PC clones to exist, after Phoenix Technologies reverse engineered the original IBM PC BIOS. The two-step process - binary to specification to unencumbered code - will no longer takes a large number of programmers and years of development time. With an LLM, it's now a thing you can do in a day. Ubiquity will make it impractical to prosecute all the people who might skip the specification step. Source code can also be covered by patents. You can be able to see their source code for a patented technique and not be able to use it without violating the law. Linux has evolved practices for dealing with this problem - one is not shipping patented video codecs, but requiring you to download them as plugins from jurisdictions where U.S. patents can't be enforced. This presents patent holders with the impractical challenge of individually suing millions of end users, assuming it can even figure out who they are. To date, AFAIK, this has not been attempted. And that's about it. If there are any other ways left to retain software secrecy and lock-in than SaaS or patents, I can't think of any. Well, you could epoxy-pot a firmware ROM, I suppose, but that can be defeated with a heat gun and a dental pick. Device manufacturers will learn not to pay for an assembly step that has become useless. Forced migration from shipped binaries to tied cloud services will be tried - Adobe pioneered this, and Microsoft is pushing it as hard as they can given that their OS is a binary that has to run locally. Both companies are seeing massive user revolts over this. There is good reason to doubt that it's a strategy that can hold customers in the long term. Also, a lot of things can't safely be tied to the cloud at all, because they can't tolerate a random network outage. Machine tools, medical devices... A whole lot of proprietary software business models are going to collapse. Hard. One that I think might survive is tax software; being able to decompile it doesn't necessarily do you a lot of good because its actual value is tracking a ruleset that changes over time and has to be maintained by vendor specialists. But cases like this are unusual. I think some dirty laundry is going to get aired, too. It has long been rumored that the reason graphics card manufacturers are so stubborn in their secrecy is that they've all been committing massive intellectual-property theft on each other for decades. If this is true, it will be exposed, and the lawsuits will be entertaining. We're entering a new world, with a lot of ancient comfortable assumptions being blown up. It's going to be fun to watch. 2/2
107
201
24
2,770
55,974
Introducing a complete map of the male fruit fly brain and central nervous system, created in partnership with @HHMIJanelia. With over 166,000 neurons, this is the largest cellular brain map to date. Learn more: goo.gle/4qQTtIW
108
582
146
5,443
811,371
Jacob retweeted
iiiiiiIIIT'S TIIIIME we proudly present: solidSF v1.0 hardware engineers, this one is for you a platform of platforms, amplifying or supporting all of the tools you love and hate folded into a clean mesh of cohesive hardware engineering excellence so what can it do?
19
15
4
209
232,584
Introducing Prime Agent: A self-improving RLM harness for coding and long-running autonomous tasks. Designed to be both token-efficient and expressive through programmatic tool calling, context as a variable, multi-agent messaging, and a self-modifiable harness state.
397
824
577
8,159
3,230,937
right monitor is 20 codex instances. left monitor has situational awareness on autoscroll. center monitor is my word doc mainfesto. two keyboards, one for both hands. left airpod is dwarkesh x eric jang, 3x speed. right airpod tchaikovsky. meta quest 3 overlays my HUD: heart rate, words per minute, blood caffeine content. one assistant hooks me to an iv of chinese peptides, cocktail. the other feeds me kimchi. my unitree robot steps in when my posture slouches. blue light beams down on me in my herman miller chair. efficiency. no wasted movement. no wasted thoughts. think you can keep up with me? good luck. this is just for my morning emails.
74
229
32
3,226
261,457
Two weeks ago, watching Agnessa Pedersen mind control a drone in real time, was one of the most moving moments in my career. Agnessa is a rare and wondrous human working towards a wild future.
108
498
106
3,625
270,090
After probably 100 hours I can finally say I am coming to a stop. Shout out to @Replit . This started for their 10 year anniversary buildathon and I haven't stopped since. Long days, sometimes 16 hour days at the weekend. I'm at a stage where I can finally sign off. KinetiCAD is browser-native parametric CAD. Real B-rep geometry kernel, OpenCascade compiled to WebAssembly. Live rigid-body physics with Rapier. Sketch, extrude, revolve, fillet, chamfer, boolean operations. Five mate types, motorised joints. An eight-material library with density-driven mass. STEP import, export and round-trip. WebGPU rendering. No install, no licence, no CAD seat. Verified by maths, not vibes: A motorised windmill assembly tuned to hold exactly pi radians per second. Across 420 simulation steps it held between 3.1415925 and 3.1415930. Target pi is 3.1415927. Every sample within 1.5e-7. Material mass is real arithmetic, not an estimate. A 5.70 cm³ part: steel at 7.87 g/cm³ gives 0.045 kg, brass at 8.50 gives 0.048 kg, PLA at 1.25 gives 0.007 kg. Exact, every time. If the maths is wrong, it is not a CAD tool. So the maths gets checked. FREE to use and open source. MIT licensed. I will always try something, even if I fail, because that's my mindset. kineticad.co.uk/ @ThreeJS #CAD #BuildInPublic #AdeviousAI
11
34
2
229
16,897
I keep trying to explain to people, but America doesn’t have a viable 3D printer company because we don’t have any of the underlying architecture. Shocking how many people think a complex product is some singular thing you build at a “factory.” A 3D printer is really just a featherweight CNC mill, so much so that they speak yeh same language (G-Code). You need: - micron accurate linear rails - Sub micron bearings for those linear rails. - High resolution servo/stepper motors. - Motion control chips and boards and stuff. - Sheet metal, injection molding, various coatings. America has *no* industrial base for *any* of this at consumer product scale. We invented or definitively innovated everything above, then the McKinsey set came in and convinced us to ship it all to China for the last 40 years. So yea, sure… you can screw a 3D printer together in America, at great expense… but you are doing so with primarily Chinese components. What we need is a bunch of uninvestible (by venture funds) businesses that are boring and competently supply a bunch of even more boring, medium margin, no-moat components.
This is what I keep coming back to. So many companies in the US and Europe, driven out of the printer business over the last few years. Each one was its own thing, there was a place for everyone, and we were all friends sharing innovations across the industry. Now everything is just a copy of a copy of a copy 📉 I'd do anything I can to make that possible again. I miss a lot of my friends from this business.
115
500
62
4,379
486,152
Introducing step . parts, an open source directory of 12,000+ STEP parts for your next CAD project Many CAD tools like Autodesk have huge closed libraries of standard mechanical and electrical parts The goal is to build an equivalent catalog for the open source CAD community
49
186
19
1,796
97,099
Jacob retweeted
What happens when you post a real Monet and say it’s AI? The coolest art social experiment I’ve seen in a while. Thank you @SHL0MS
950
3,249
1,123
20,624
2,284,149
AI designs functional hardware! 🔧 Claude Opus 4.7 is leading CAD benchmarks, and AI is starting to design functional hardware autonomously. The shift happening: CAD as a tool → CAD as an autonomous system. The model didn't just sketch geometry, it generated mechanically valid designs with joints and motion constraints. It optimized for real-world functionality, not just visual correctness like earlier LLMs. This mirrors what happened in coding last year. In software, engineers moved toward requirements, architecture, and orchestration while code became the compiled outcome of AI systems. Mechanical engineering may follow the same path: from designing parts → defining intent. This changes the role of engineering teams. Instead of manually designing every component, engineers will define constraints, validate outputs, and focus on system-level architecture. Engineers spend less time on geometry and more time on optimization and integration. AI handles the mechanical validity checks that traditionally required experienced designers. ~~ ♻️ Join the weekly robotics newsletter, and never miss any news → ziegler.substack.com
12
55
5
410
28,260