frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

Show HN: Latent-free ternary LLM training

https://github.com/ValerioDolci/bitbop
1•Feyd•41m ago

Comments

Feyd•40m ago
In the last week I have been researching LLM architectures, putting together all the known ways to train AIs and ended up creating something apparently new. Long story short, the vRAM needed to train my LLM seems to be up to 5x times smaller than a standard AdamW. I was studying the ternary weights on my own when I asked myself if it was possible to actually train an LLM with this method. Turns out, yes, it is. I created a POC about the possibility to train a latent-free LLM, significantly saving memory for training, and shrinking the shipped model size for storage/distribution (inference speed/RAM need a dedicated kernel — not built yet. During tests on a toy scale on babyLM benchmark (up to ≤125M), performance proved to be on par with other autoregressive models. As of today, there is no optimized kernel, so actual vRAM savings potential and speed-up are not completely fulfilled. This is clearly still a WIP, as I am experimenting with other configurations and scale-ups. If you are curious about this, the repo is open with more technical information. https://github.com/ValerioDolci/bitbop

A Go daemon that matches Tesla charging to solar export (Sungrow and Tessie)

https://github.com/guiguan/tesla-sungrow-charge-controller
1•guiguan•53s ago•0 comments

Grok uploaded my user directory to xAI's servers

https://twitter.com/a_green_being/status/2076598897779020159
2•tnolet•1m ago•0 comments

Card-type tool to support divergent thinking to embody benefits of inconvenience

https://www.researchgate.net/publication/279855476_Card-type_tool_to_support_divergent_thinking_f...
1•andsoitis•4m ago•0 comments

Grok CLI uploaded the whole home directory to GCS

https://twitter.com/i/status/2076598897779020159
3•denysvitali•6m ago•0 comments

Lifestyle interventions for the prevention of cognitive decline in older adults

https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(26)01278-X/abstract
1•wslh•6m ago•0 comments

Is there any reason to think dolphins and humans have a special relationship?

https://aeon.co/essays/do-dolphins-really-share-a-special-bond-with-humans
1•prmph•8m ago•0 comments

Show HN: Free on Device Meetings Summaries Right to Apple Notes

https://voice.baselinemakes.com/
1•stevenhubertron•8m ago•0 comments

Show HN: Nable, open source finops MCP

https://github.com/chaandannn/finopsmcp
1•piosin•8m ago•0 comments

Sun Microsystems CEO: You have zero privacy anyway. Get over it

https://www.techradar.com/pro/quote-of-the-day-by-sun-microsystems-ceo-scott-mcnealy-you-have-zer...
1•robtherobber•9m ago•1 comments

Users report that GPT-5.6 Sol has become less capable than its initial release

https://twitter.com/i/status/2076495156757577895
1•maxloh•9m ago•0 comments

Leave a failing test before you go on vacation

https://lukapeharda.com/article/leave-a-failing-test-before-you-go-on-vacation/
1•surprisetalk•10m ago•0 comments

No File, No Path, Still Running: Inside Linux Fileless Execution

https://mrn1gativ3.github.io/posts/filelessexec/
1•mashally•10m ago•0 comments

Stop falling for the "But China" excuse [video]

https://www.youtube.com/watch?v=qde4r4_uMU0
1•leonidasrup•11m ago•0 comments

AI and Job Postings: From Destruction to Creation

https://www.hiringlab.org/2026/07/08/ai-and-job-postings-from-destruction-to-creation/
1•hartleybrody•16m ago•0 comments

Madame Semver Will See You Now

https://nesbitt.io/2026/05/10/madame-semver-will-see-you-now.html
1•progbits•16m ago•0 comments

If you want to make money, be useful to rich people

https://filiph.net/text/to-make-money-be-useful-to-rich-people.html
1•speckx•18m ago•0 comments

How to Build a Low-tech Website? (2018)

https://solar.lowtechmagazine.com/2018/09/how-to-build-a-low-tech-website/
1•downbad_•19m ago•0 comments

Who Owns What Your Agent Knows?

https://www.safiware.com/p/who-owns-what-your-agent-knows
1•safiware•20m ago•0 comments

MITS: Rockets, Calculators, and Personal Computers

https://www.abortretry.fail/p/micro-instrumentation-and-telemetry
1•BirAdam•21m ago•0 comments

Ask HN: Active Forums You're a Part Of

2•insanejules•23m ago•0 comments

Logical Query Processing (2009) [pdf]

https://accessexperts.com/wp-content/uploads/2015/07/Logical-Query-Processing-Poster.pdf
1•rzk•23m ago•1 comments

Pactiamo|Proposals, contracts, signatures and payment in one link

https://www.pactiamo.com
1•kkmigu•25m ago•0 comments

The State of AGENTS.md: scoring the 16 biggest AI agent repos' own instructions

https://fpaul.dev/writing/state-of-agents-md-2026/
1•Fpailo•25m ago•1 comments

L8ER SK8ER, a history of a book that never sold

https://biblicallyaccuratealiens.substack.com/p/l8er-sk8er-a-history
1•meandthemajor•26m ago•0 comments

The New York nurses replaced by AI: 'It should concern every patient who cares'

https://www.theguardian.com/technology/2026/jul/13/nurses-new-york-ai
3•beardyw•27m ago•0 comments

A Right Too Expensive to Exercise Isn't a Right

https://deepbluedynamics.com/blog/right-too-expensive-to-exercise
3•kordlessagain•33m ago•0 comments

69% support forcible transfer of 50% AI stock to public fund: Verasight

https://reports.verasight.io/reports/june-2026-ai-survey
2•akramachamarei•33m ago•0 comments

Show HN: ProtoLink – A Python runtime for distributed A2A agents

https://github.com/nMaroulis/protolink
2•nmaroulis21•35m ago•0 comments

Show HN: A terminal browser you can use over SSH

https://duetbrowser.com/krnl-ss-redirect
1•keepamovin•35m ago•1 comments

How the most impactful AI startups will be built in emerging markets

https://blogs.worldbank.org/en/voices/small-ai--big-bets--how-the-world-s-most-impactful-ai-start...
1•alphabetatango•37m ago•0 comments