frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Built a site that teaches investing using the live market as the examples

https://www.stocksightfinance.com
1•kshitik•15s ago•0 comments

I let Claude review the C code I wrote as a math intern in 2013

https://mitsudomoe.github.io/Kinetic-Solver/
1•MathLab•1m ago•1 comments

Ole Bielfeldt: Finding the Extraordinary in Everyday Objects

https://www.nikonsmallworld.com/masters-of-microscopy/ole-bielfeldt
1•smusamashah•2m ago•0 comments

Show HN: One REST endpoint to filter 80+ financial datasets

https://financialdata.net/universal-query
1•_FDN_•3m ago•0 comments

Max is now source available and open to everyone

https://www.modular.com/blog/max-license
1•lairv•5m ago•0 comments

Ask HN: Secure minimalistic web front-ends to Local LLM's?

1•Bender•5m ago•0 comments

Nvidia Sentry: Secure AI Agents at Runtime

https://developer.nvidia.com/blog/inside-nvidia-sentry-secure-ai-agents-at-runtime/
1•hasheddan•5m ago•0 comments

Judgment Becomes the Job

https://twitter.com/ahatamiz1/status/2113344049784799378
1•pretext•6m ago•0 comments

How to Unclench

https://howtounclench.com/
1•surprisetalk•7m ago•0 comments

Superhuman Hardware Design Is the Easy Part

https://www.a16z.news/p/superhuman-hardware-design-is-the
1•gmays•8m ago•0 comments

Bolsonarism has formed its own digital party, says Marcos Nobre

https://jornal.unicamp.br/en/noticias/2026/06/26/bolsonarismo-constituiu-seu-partido-digital-afir...
1•cosmophany•8m ago•1 comments

AI Article Reviews Dominate Human Peer Reviews at Their Best [pdf]

https://gking.harvard.edu/files/peerless.pdf
1•mpweiher•9m ago•0 comments

I Built an AI Bot to Fact-Check the Midterms. Here's What I've Learned

https://agentsofinfluence.substack.com/p/i-built-an-ai-bot-to-fact-check-the
2•laurex•12m ago•0 comments

Artifact Arena: 'Making' is the measure of intelligence

https://artifactarena.ai
1•tbruckner•13m ago•0 comments

Ask HN: Is the downside of investing in AI real?

2•okzgn•13m ago•0 comments

Pentagon grants DHS permission to test heat ray weapon for use at border

https://www.cnn.com/2026/10/21/politics/heat-ray-weapon-dhs-pentagon
1•bushwart•14m ago•0 comments

Trust Stores Are Dependencies Too

https://nesbitt.io/2026/10/21/trust-stores-are-dependencies-too.html
1•keks0r•15m ago•0 comments

eMudhra ignored Mozilla Root investigation deadline; blames social media

https://bugzilla.mozilla.org/show_bug.cgi?id=2083449
2•IcePenguin•15m ago•2 comments

Officials confirm flesh-eating New World screwworm parasite found in US wildlife

https://www.theguardian.com/us-news/2026/oct/21/new-world-screwworm-texas
1•thunderbong•16m ago•0 comments

Hex Map 5.7 – Set Terrain Type Burst Job

https://catlikecoding.com/unity/hex-map/5-7/
1•ibobev•16m ago•0 comments

Trailing SOTA Is a Feature for Profit Maximisation Under Compute Constraints

https://twitter.com/jenzhuscott/status/2113108957355229513
2•ibobev•17m ago•0 comments

Postgres 19 is late, and it's AI's fault

https://neon.com/blog/postgres-19-is-late-and-its-ais-fault
1•nikita•18m ago•0 comments

First-of-Its-Kind Vaccine Ushers in New Hope for Cancer Patients

https://www.nytimes.com/2026/10/23/well/cancer-vaccine-melanoma-study.html
2•gumby•19m ago•0 comments

Tell HN: Unbounce Security Incident

2•coderintherye•19m ago•0 comments

New discoveries, old knowledge: AI finds connections humans missed

https://gavinray97.github.io/blog/undiscovered-public-knowledge
1•gavinray•19m ago•0 comments

MemCurrent: Designing Persistent Memory for Production Agents

https://www.databricks.com/blog/memcurrent-designing-persistent-memory-production-agents
1•gmays•20m ago•0 comments

When the Term "Heatwave" Is No Longer Enough

https://coveringclimatenow.org/from-us-story/when-the-term-heatwave-is-no-longer-enough/
4•robtherobber•20m ago•0 comments

An alignment assessment of the UK AISI cybertesting incident

https://www.anthropic.com/research/alignment-assessment-ukaisi
1•olwmc•22m ago•1 comments

PGSimCity Turns PostgreSQL Complexity into a Virtual City 3D Simulation

https://www.infoq.com/news/2026/08/pgsimcity/
1•woliveirajr•22m ago•0 comments

MemoryTax: The Bitter Lesson of Agent Memory

https://arxiv.org/abs/2610.29823
1•netsroht•22m ago•0 comments
Open in hackernews

Ask HN: How to Foreshadow?

2•ksec•1y ago
Are there any way where I can foreshadow certain things and only have it revealed at a certain date? With Proof this was written way before its current happening time?

Comments

codetrotter•1y ago
Make a PDF or whatever kind of document containing the things you want to reveal.

Make a hash sum of the file.

Publish the hash of the document somewhere. Like on your Mastodon timeline or on Reddit, or on HN, or as an ad in a national printed newspaper.

Make sure the text is pretty long though. If you write something super short someone could guess the hash for it.

Write a short comment along with the hash. Something like:

“My prediction will be revealed on Dec 31 2025. Sha256 a9774a1a6ebf564cc408cfd86b5f2c06c13d830e143989714d958d34f325db13.”

When time comes, publish the document.

codetrotter•1y ago
Another alternative is to encrypt the document with a very strong pass key, and publish only the encrypted document now along with an announcement that the key to decrypt it will be published on such and such date.

If you are on a platform that doesn’t support posting attachments you can use IPFS or similar to host the encrypted file.

pvg•1y ago
If you write something super short someone could guess the hash for it.

You can skip the pdf and unless the thing is so short/predictable you can guess it outright, nobody is going to 'guess' the hash. E.g. 'I made up a random 64 byte sequence, here's its cryptographic hash, guess the sequence' - that doesn't happen.

codetrotter•1y ago
If the sentence is too short it might be possible to guess.

It’s not going to consist of random bytes.

Imagine for example that a high profile celebrity accused of murder has successfully defended himself in court and on that day he tweets:

I have something to reveal, but I won’t be telling you until the year 2135. Sha256 73fa194342af6b6e355e129fafd9b19d2a63589b1e3c6e2b5d94a7ca1b3e25f6

If this happens to be a short sentence related to recent events, don’t you think someone will figure out what he wrote pretty soon?

pvg•1y ago
You can just pad or HMAC the short sentence.
codetrotter•1y ago
Absolutely. Padding it falls under making sure the text is not too short.
gus_massa•1y ago
Bonus points if it's a platform where you can't delete the posts. Or you number the posts like

> My prediction 683 will be revealed on Dec 31 2025. Sha256 a9774a1a6ebf564cc408cfd86b5f2c06c13d830e143989714d958d34f325db13.

> My prediction 684 will be revealed on Nov 7 2025. Sha256 17147091grhef7wy89908409849rwiruriwu87w980241989171n109288021982.

It would even be better to add it to a blockchain to ensure there is no backdating.