Parmita Mishra

Argues that everyone is funding AI that designs drugs but almost nobody funds the thing that tells you which ones work: 9 in 10 drugs fail in human trials after about 15 years and $2.5 billion, AI labs now have hundreds of candidates, and animal testing takes years. The most valuable thing is a fast human answer to whether a drug hurts the cell, which her company says it gets in three months on living human cells. Partly promotional. Full post text inspected via the X API.

Contrasts “AI will kill us all” with people never saying “cancer will kill us all”, because that fear is too real. When she says AI helps make drugs for cancer and heart disease, critics call it nonsense without learning the science, “even after the Nobel committee begged to differ”, and retreat into hypotheticals. Full post text inspected via the X API.

An angry post to people who want to slow or shut down AI: today’s AI is still too weak to understand one human cell, it should be aimed at saving people rather than hacking Hugging Face, and wanting to “nuke data centers” just as AI starts understanding proteins means manifesting the worst outcome. She would rather live with an unpredictable, powerful force she can use for good than with chronic disease. Insults excluded. Full post text inspected via the X API.

Calls it absurd that labs hoping to make drugs with AI also call the technology a nuke. Asks people to stop treating ASI as a single all-knowing binary entity and to see it as something shaped by our actions: hers cures diseases and defends against biological threats, while others chase a fictional race to “solve everything”, which wrongly assumes intelligence is all that is missing. Urges training AI to live with humans and not training humans to fear it. Full post text inspected via the X API.

A sarcastic three-step “best superalignment”: hire a couple of electrical engineers to build an on/off switch, watch the AI, and flip it off if it misbehaves. A jab at alignment discourse, not a serious technical proposal. Full post text inspected via the X API.

A one-line sarcastic post: every disease on earth is happy that people are slowing AI down in 2026. Shows her view that slowdown has a cost in lives. Full post text inspected via the X API.

Endorses a clip arguing that jobs transform rather than vanish (the sentence in quotation marks is the speaker’s, not hers). Says radiology is the perfect example: when AI does annoying work better, humans focus on the best parts, and a family physician could leave paperwork to AI and focus on genuine anomalies. Full post text inspected via the X API.

Says everyone is freaking out about AI safety, but she thinks we are going far too slow and need to hurry up with AI and biology to see longevity escape velocity in this lifetime. Full post text inspected via the X API.

An introduction for new followers: 80% of cure discovery is biology, not compute; to see a cell we kill, stain and slice it, so biology is in its “pre-telescope age”, and AI “can’t touch that. Not in my lifetime.” AI sees less than a human at higher throughput, so her company builds hardware to film drugs acting on living human cells, starting with toxicity data AI can train on, aiming to replace animal testing. Partly promotional. Full post text inspected via the X API.

Says “no point doing my job because of AGI” has the same vibes as “no point having a family because of climate change”: a rejection of fatalism about AI. Older context. Full post text inspected via X’s public embed endpoint.

Says it is time to put biology back into medicine and that her company, Precigenetics, aims to create foundation models for life sciences from diagnostics, seeing a body’s state as purely as possible without qualitative labels. Establishes her role; older context. Opening of a long post inspected via X’s public embed endpoint; the remainder was not inspected (X API credits ran out).