HNHacker News
TopNewBestAskShowJobs

ihodes

4,397 karma · joined March 19, 2010

Building foundation models, helping to advance science

isaachodes@gmail.com

submissionscomments
ihodes··on Show HN: Krabby – A keyboard interface to the web, inspired by Kakoune
Is there something like this that's currently developed for Safari?
ihodes··on Ask HN: What's the most valuable thing you can learn in an hour?
This depends on how the eggs are processed prior to you purchasing them, so beware. Cf. https://www.npr.org/sections/thesalt/2014/09/11/336330502/wh...
ihodes··on The insulin racket: why a drug made free 100 years ago is recently expensive
They do that, or contract with companies (like PBMs) to do so. That's been the way this works forever. Asymptotically no one pays the ~$40k+/mo list price.
ihodes··on Retail stores use Bluetooth beacons to track customers
That's really awful; I'm so sorry you both had to go through that.

Would you mind sharing more information about how you found that list (esp for a given zip), and how you think they tied that information to an address? My email is in my profile, if you wouldn't mind reaching out.

ihodes··on Ask HN: How do you organize/track your personal goals?
Might be of interest to you https://sloanreview.mit.edu/article/when-smart-goals-are-not...
ihodes··on Universal Basic Income Is Not Feasible
Check out Kent Smetters if you're curious.
ihodes··on Finland Has Second Thoughts About Giving Free Money to Jobless People
No, they explicitly made this a conditional handout this year, which is very much a termination of the original plan. UBI researchers have generally considered this a failure already (not to mention that many economists think UBI cannot work, anyway).

"Finland has actually reversed course on that front this year, adopting rules that threaten to cut benefits for jobless people unless they actively look for work or engage in job training."

ihodes··on Finland Has Second Thoughts About Giving Free Money to Jobless People
I believe it was supposed to run for two years. They ended it a year early. (First source that pops up: https://www.cnbc.com/2017/01/03/finland-experiments-universa...)
ihodes··on Medicare will require hospitals to post prices online
Thank you!
ihodes··on Medicare will require hospitals to post prices online
I'm unable to see that in your profile (looks blank to me). Do you mind updating it?
ihodes··on Medicare will require hospitals to post prices online
I'm curious to learn more about your clinic. Mind sharing any information?
ihodes··on San Francisco’s Skyline, Now Inescapably Transformed by Tech
The size comparison is a bit off; Manhattan is about half the size of SF (https://www.comparea.org/r2552485+r111968). Of course, NYC is over 6x the size of SF (https://www.comparea.org/r175905+r111968).
ihodes··on Alphabet's Sidewalk Labs to turn 800 acres of Toronto into an “internet city”
Also seems like a useful captive waypoint to monitor MAC addresses & movement across a city.
ihodes··on Some insurers insist that patients forgo generics and buy brand-name drugs
There are studies showing that some generics don't perform as well as the brand name. Different formulations of the same drug (e.g. extended release medications, different sized particles in inhalers, etc) can have vastly different effects.
ihodes··on A vigilante trying to improve IoT security
This is a more in-depth source for the same story: https://arstechnica.com/security/2017/04/brickerbot-the-perm...
ihodes··on Why ML/OCaml are good for writing compilers (1998)
What does the lack of multi core keep you from doing?
ihodes··on The findings of medical research are disseminated too slowly
For those not in the field, it's worth noting that a good deal of research is disseminated prior to publishing (or while "in press"/waiting to be published) via conferences or word of mouth.
ihodes··on Adventures in /usr/bin and the likes
In emacs, `C-u M-| cmd <enter>` does this.
ihodes··on Apache Kudu – Fast Analytics on Fast Data
I have definitely have shared that sentiment. However, Apache didn't create this project; it's from Cloudera originally, and they've got a pretty good track record. In this case, Kudu is more of a database/store on top of which analytics products (for example, this is a best-in-class engine for Impala, allowing fast read/write) can run. I wouldn't quite say it's like HDFS, which is more of a document store/distributed file-system.
ihodes··on 1 Patient, 7 Tumors and 100B Cells Equal 1 Striking Recovery
Took a few notes when reading the paper[1], expanded some acronyms as well. Definitely a nice result, and an interesting paper.

* Steve Rosenberg's a respected expert in the field

* Adoptive T cell therapy (ACT) of TIL (tumor-infiltrating lymphocytes) T cells specific for a (common) KRAS mutation (G12D)

* Culture/expand TILs, use tandem-minigenes to narrow down most reactive (to in silico ID mutations) TILs, then use synthetic peptides on the mutations encoded in the TMG that elicits largest response to find which epitopes in particular were being reacted to.

* 6 tumors eradicated, 7th shrunk but recurred (had to be resected, patient has no disease 4 months later); 7th had lost the HLA-C allele that the TCRs were specific to

* The supplementary materials/methods were pretty interesting (very technical)

[1] T-Cell Transfer Therapy Targeting Mutant KRAS in Cancer http://www.nejm.org/doi/full/10.1056/NEJMoa1609279

ihodes··on Cells edited using CRISPR–Cas9 injected into a person for the first time
I haven't read a ton of CRISPR/CAS papers, but I have read the seminal ones, and didn't see that. Mutations in unmodified cells are completely expected, of course, but not mutations in line with where CRISPR is being targeted—I'd hardly believe that to be the case. It looks like the cell perforation (getting CRISPR/CAS into the cells) can certainly affect cell viability, but those numbers have been steadily, rapidly improving. Quick search shows that in 2015, a study found 80% viability post perforation and CRISPR/CAS application (http://advances.sciencemag.org/content/1/7/e1500454.full).
ihodes··on Cells edited using CRISPR–Cas9 injected into a person for the first time
It's certainly the former; CRISPR is used to mess with a gene, not kill a population of cells. This isn't what the trial is looking to validate.
ihodes··on Cells edited using CRISPR–Cas9 injected into a person for the first time
Startups working on this (with credible founders, i.e. inventors of CRISPR) include Editas Medicine, Caribous Biosciences, and CRISPRx. Look out for Doudna, Qi, Charpentier, Elledge, and Church. Also see J. Listgarten for computation approaches to improving CRISPR/CAS9.

This isn't ready for in vivo use yet due to off-target effects (CAS9 is a very permissive protein that often binds to the wrong part of the DNA).

In the US, Carl June's group at U Penn is working on editing adoptive, engineered CAR T cells for cancer therapy, going into clinical trials. They're probably bummed to have been scooped (first to do an adoptive CRISPR engineered T cell transfer), though their work is certainly more clinically relevant than Lu's.

ihodes··on Why Medical Bills Are Killing Us (2013) [pdf]
Your questions are very good, and I answer them below briefly in the interest of expediency, not to be rude.

1. Some knowledge is shared in journals; most knowledge isn't published until years after it's established. Much is never published, and is learned through word of math, and institutionally encoded. MD Anderson is world-renowned for being a center for excellence in cancer care.

2. Doctors don't read every article published in every journal related to their work. Even if they did, or read much of it, synthesizing into practice is a very different (and difficult) matter. Many of them don't read at all.

3. Why would MDs want to be send worldwide? And why would MD Anderson want to send them all over? Who pays for this? How can a doctor affect change in a large hospital system?

ihodes··on Ask HN: Who is hiring? (November 2016)
Hammer Lab | New York, NY (NYC) | Full-time, onsite

We’re a team of software developers and data scientists working to understand and improve how the immune system battles cancer. We're looking for (1) a senior machine learning specialist and (2) a senior development operations engineer.

1. Both cancer and the immune system are complex collections of cells whose interactions are difficult to predict. If you have done research in a top deep learning lab and have built and deployed models used in a high-value application, your skills could improve outcomes for patients enrolled in clinical trials of personalized cancer vaccines and combination immunotherapies.

2. The ideal candidate will have installed and serviced server and network hardware in a data center, written software to install, configure, and monitor production services, structured the build and release process of high-quality open source software, and secured and governed petabytes of data. We recognize that it’s rare to find an individual with experience in all four of these domains, however, so we are happy to hear from anyone with strong skills in one or two domains and an interest in learning more about and having responsibility for the rest.

Check out the job descriptions for more information, and send us an email! http://www.hammerlab.org/jobs/

ihodes··on Ask HN: Do you get paid to contribute to a open source project?
Feel free to email me if you're interested (even if you don't feel as though you're an ML expert)!
ihodes··on Ask HN: Do you get paid to contribute to a open source project?
(Nearly) all our work (github.com/hammerlab) is OSS. We're hiring experts in ML if you're interested in working on big genomic data in the field of cancer immunotherapy. We're a lab (academic not-for-profit, part of Mount Sinai's medical school, run off foundation, grant, and gift money) of software engineers from industry and academia in NYC trying to make research better, and cure some cancer while we're at it (running some clinical trials).
ihodes··on Tesla plunges in Consumer Reports' rankings
Buying shares of TSLA doesn't put money in TSLA's pocket anymore than buying a PUT on TSLA takes money from their pocket.
ihodes··on Introduction to Functional Programming in OCaml
I used to write a fair bit of Clojure, in 2009-2011, but have moved to mostly OCaml these days.
ihodes··on Introduction to Functional Programming in OCaml
You don't need to leave monads in OCaml (in fact, I'd be very sad to not be using them regularly). Typically the Lwt, Option, and Error monad. (Also speaking as a former Clojurer).
← PreviousPage 2 of 14Next →